In late March, I posted on Twitter that the NIL market was up about 35% from last season based on proprietary data I have at EvanMiya.com. That number is already very outdated. Based on input from teams, the NIL market for Division 1 players is up about 65% from last year. That means that a player good enough to be valued at $1 million last offseason would be getting around $1.65 million this offseason.

In today’s article, we are going to examine several key trends in the NIL market:

• NIL market changes from last offseason

• Market demands for high-major vs mid-major players

• Player types that are being overvalued in the current market

Where The Data Comes From

Before I summarize the most interesting trends in the college basketball NIL market, it’s important to explain why I can do this analysis to begin with. Since there is no publicly verified database for player contracts, team budgets, or NIL rates, very few people can reliably convey knowledge of the market without making some big assumptions.

About a year ago, I launched a tool called the “Front Office Suite”, which is specifically designed for coaches, GMs, and ADs to obtain NIL evaluations for players, based on their analytical profiles, while being responsive to the current market trends. Because the NIL market is constantly changing, it’s crucial for any tool that attaches dollar figures to players to be flexible enough to adapt to market trends.

I’ve conducted extensive research on how numeric player value correlates with monetary value, drawing on salary cap science and the impact of individual players on team success. While I can reasonably tell somebody that Player A should be worth 25% more NIL dollars than Player B, it would be a stretch for me to assign fixed NIL figures to each player without allowing for adjustments to the fluid market.

Here is a brief summary of how Front Office Suite NIL evaluations are calculated, leaving out a few proprietary details:

Each team enters NIL values for individual players that reflect the current market rates they are seeing in recruiting. Each NIL number entered for a player doesn’t have to be exact amount they will likely make, but when you pool 10-15 of these data points together, it does a pretty good job of summarizing where the overall NIL market is at today. Data entered stays private to each team’s profile and is not shared across users. While there is no way to verify the accuracy of the data entered, there is no incentive for a user to lie or fudge the numbers, since more accurate intel leads to more realistic NIL approximations in the Front Office Suite. Once these data points are entered, a user can then see the true “NIL value” for every D1 player based on each player’s projected value next season, their analytic profile, and where they should be priced according to market trends. Though each team sees slightly different NIL evaluations based on the intel they provide to the tool, most teams end up with very comparable dollar figures after putting in 10+ data points. Users can add or modify any of these settings as they see fit throughout the year, allowing them to stay up to date on current market trends and see more accurate NIL estimations for all players.

The NIL Market Is Way Up

There are about 30 teams that use the Front Office Suite, spanning 12 conferences in the high-major and mid-major ranks. By summarizing the NIL player data entered across all teams and comparing it to last year’s market data, we can gain a pretty good understanding of how the NIL market has changed over the past year.

As mentioned at the top, the NIL market for Division 1 players is up about 65% from last year. That means that a player good enough to be valued at $1 million last offseason would be getting around $1.65 million this offseason. It’s pretty hard to fathom how market rates have increased this much, especially since we saw a similar increase from 2024 to 2025, if not even higher.

The market is up even higher for players coming from high-major programs: There has been a 73% increase in the going rate for players on power conference teams. The market increase for players from low- and mid-major conferences is slightly lower, up 44% from last offseason.

Bigs Are Being Overpaid

It shouldn’t come as a surprise that bigs are a prized commodity this offseason after teams like Michigan, Arizona, Florida, and Duke all dominated with loaded frontcourts. Michigan and Florida, in particular, thrived by having players like Yaxel Lendeborg and Thomas Haugh, traditionally considered stretch-4s, play down at small forward. Since we live in a copycat league, other teams are trying to replicate this by stocking up on extra big men.

Predictably, this has led to some pretty outrageous asking prices for bigs all across Division 1. In a recent coaches poll, Jeff Goodman approximated that elite portal bigs were going for at least a million more than guards or wings. In many cases, this overspending on bigs is not justified.

By looking at players across positions who are expected to have similar value next year according to our player projections, we can compare their actual going rates to determine market demand.

The table below shows how much it would cost a team to acquire a player worth $1 million at each position. Centers are being the most overpaid, costing about 30% more than they should on average, while point guards are more of a bargain.

Here’s how to interpret the table exactly: to obtain a center or a point guard who are equivalent in talent level, it would cost $1.3 million for the center, while only $813,000 for the point guard. That means, in this market, it costs 61% more NIL dollars to get a center of the same caliber as a point guard.

How would this market discrepancy translate to player value on the court? Without getting too into the weeds, let’s pretend that a team had $1,000,000 to spend on an extra player. For the sake of this example, let’s assume they already have rotation pieces figured out and are willing to either get a starter-level PG or C to add to the mix. Based on the high demand for centers, they could reasonably get a big with a projected Bayesian Performance Rating next season of 3.9. A player of that caliber would be a borderline top 450 player in college basketball next season. If they were to spend that same $1 million on a point guard instead, they could get a better player, with a projected BPR of about 5.1, due to lower demand in the guard market. A player of that caliber would be around a top 250 player next season.

For a team hoping to make the NCAA tournament, that difference in player caliber would be worth about a 5-spot jump in their preseason roster talent ranking! Obviously, this example is a bit contrived, but it illustrates how this high demand for bigs can significantly affect the quality of players a team can get within its budget.

I should note that the table above does not indicate that point guards are the cheapest position to acquire. An elite-level point guard in college basketball is more valuable than an elite-level wing in today’s game, which is why the cost for a quality point guard is higher. The point is that teams are still getting point guards at the biggest bargain compared to all the other positions, based on the production they will actually get out of those players.

Rosters That Can Beat The Market

With more teams trying to load up on long, athletic frontcourts, my current working theory is that teams who stock up on value at other positions could end up with more talented rosters.

Vanderbilt’s roster from this season is a great example. Jalen Washington was their only true-sized big who played regular minutes at 6’10”, and he was just their 6th most valuable player according to BPR. Their other frontcourt spots were filled by wings and undersized bigs: Tyler Nickel, Devin McGlockton, and AK Okereke were all 6’7”. At the guard spots, Vanderbilt’s best players and leading scorers were Tyler Tanner (6’0”) and Duke Miles (6’2”). Mark Byington’s squad had a ton of skill across all five positions despite the lack of size, which allowed them to be so successful.

Nebraska, which beat Vanderbilt in the Round of 32, was also built very similarly. Rienk Mast and Berke Buyuktuncel, both 6’10”, combined for 53 minutes a game. Outside of those two, the rest of the roster was 6’7” or shorter, but filled with highly-skilled players who had great chemistry together.

While I’m not claiming that the 2027 national champion will have a roster that looks like Vanderbilt’s or Nebraska’s from this year, I do think that many teams who overspend on bigs trying to chase the Michigan blueprint will end up being disappointed with the production they get from their rosters. Teams that buck the trend by seeking value elsewhere may be more successful.

Further Analysis

I will be expanding on this analysis and analyzing some of the most interesting portal moves in my next episode of The EvanMiya College Basketball Show later this week. Be sure to subscribe to the podcast platforms or on YouTube!

• Four individual region previews with Terrence Oglesby, Isaac Trotter, Jim Root, and Brian Rauf

• Title Favorites

• Under-seeded teams + bracket value picks

• Kill Shot analysis

• Teams with high ceilings and low floors

… and more.

You can find it on audio podcast platforms like Apple and Spotify here.

The video version on Youtube has all of the charts up on the screen so you can follow along:

In this bracket primer, we will discuss:

• Who are the true title favorites?

• Which teams have the most value based on where they are seeded?

• Who are this year’s cinderellas?

• What can the “Kill Shot” tell us about tournament success?

• Which teams have high ceilings and low floors?

• Which teams are the most consistent game-to-game?

Without further ado, let’s dive in!

Predicted Efficiency Landscape

The place to start is by examining each tournament team’s predicted strength at EvanMiya.com and comparing offensive and defensive efficiency. These ratings are adjusted for current injuries, which means teams like Texas Tech (JT Toppin), Duke (Caleb Foster, Patrick Ngongba), UNC (Caleb Wilson), and BYU (Richie Saunders) are appropriately docked for the uncertainty surrounding their players’ availability. You can find complete injury reports, updated daily, in the Matchup Previews tool at EvanMiya.com.

The graph below shows every team’s predicted offensive rating and defensive rating, with teams split into tiers based on their tournament potential:

Title Favorites

As mentioned in a previous article, teams heading into the tournament in the “Title Favorites” tier have won the last 10 national titles. This year, eight teams qualify: Michigan (1 seed), Duke (1), Florida (1), Arizona (1), Iowa State (2), Houston (2), Purdue (2), and Illinois (3).

The four 1-seeds, in particular, are strong again this year. According to bracket simulations at EvanMiya.com (at the bottom of the article), there is a 78% chance that a 1-seed will win the title, which is even better than the actual track record of 1-seeds. Since 2000, 18 of the 25 national champs (72%) have been 1-seeds. There is a 9.1% chance that we get consecutive Final Fours with all 1-seeds.

Tournament Team Strength vs Bracket Seed

To better understand how these team ratings compare to each team’s seed, I present the graph below, which highlights teams that are much stronger or weaker than usual for their seed. This is really where we can find value in the predictive analytics. Teams highlighted in orange are considered “underseeded” and advance 69% further than other teams with the same seed.

This year, 13 teams qualify as underseeded because they are much better than their seed would suggest. Of these teams, the ones that really stick out are 5-seed St. John’s, 6-seed Tennessee, 11-seed Texas, and 13-seed Hofstra, all of which have a predicted team strength much better than their seed.

This year, the group of 11 seeds is particularly strong relative to other teams around them. VCU is favored to beat 6-seed UNC in their first-round matchup, and the winner of Texas and NC State will be in a pick-em game versus BYU in the first round. Ohio State was always going to have a tough time getting to a Sweet 16, but if there’s any 8-seed to it this year, it’s the Buckeyes.

St. John’s has the best chances of advancing to the Final Four (8.9%) of any team not in the top three seed lines. If they advance past Northern Iowa, they would face the winner of Kansas/Cal Baptist, with a likely Sweet 16 matchup against Duke after that.

On the flip side, some overseeded teams are likelier to come up short in the tournament than usual expectations for their seeds: 5-seed Texas Tech (JT Toppin out), 6-seeds BYU (Richie Saunders out) and North Carolina (Caleb Wilson out), and 8-seed Villanova.

The Kill Shot

This wouldn’t be an EvanMiya.com article without analyzing the “Kill Shot”, any time a team scores 10 or more unanswered points in a game. Teams that go on many of these 10-0 scoring runs are more able to erase a deficit quickly and can also easily close the door on an opponent in a tight contest. Similarly, teams that limit opponents’ “Kill Shots” are more likely to hold on to a lead rather than allow another team back into the game.

Before we jump into analyzing individual tournament teams, here are the stats about scoring runs that you need to know:

• A team with at least one 10-0 run wins 71% of the time

• A team with more Kill Shots than the opponent wins 81% of the time

• A team with 2 Kill Shots in a game wins 88% of the time

• A 15-0 run leads to a win 86% of the time

• A 20-0 run leads to a win 91% of the time

The graph below shows the landscape of tourney teams in 2026 in terms of scoring runs per game and scoring runs conceded per game, adjusted for opponent quality:

As shown on the graph, I like to put teams in four categories: “Strong” teams, “Streaky” teams, “Least Streaky” teams, and “Suspect” teams.

All the major title favorites fall in the “Strong Teams” category, but Saint Mary’s and Nebraska are here as well.

Michigan is the best team overall at going on 10-0 runs and not giving them up in return. Meanwhile, Arizona has only given up a single Kill Shot in a meaningful game all season.

Illinois has the best adjusted Kill Shot rating of any team, meaning that they are more likely to generate a 10-0 run than any team in the field.

This is where we can really have some fun with this metric. The nine tournament teams in the “streaky” category are all very capable of going on scoring runs… but they also give them up just as quickly. Games with these teams are often wild, with quick lead changes back and forth. No game with a “streaky” team is over, win or lose, until the very end.

There are two main takeaways for teams in the “streaky” category. First, they are vulnerable as a favored team but have greater upset potential when they are the underdog. Because scoring spurts can happen at any time in either direction, their games are very unpredictable. A team like Iowa State could either flame out in an early round or get hot and go deep in the tournament. And you really don’t know which one will happen.

The other takeaway is that live spreads for games involving these teams could be less accurate, as any lead could be vaporized quickly. If a team like Saint Louis is down 10 at halftime, they will be more likely to mount a comeback than most other teams of similar quality.

In contrast to the last category, we now look at teams in the “least streaky” group, which are more boring by nature. These teams have not had as many 10-0 runs this season, but they rarely give them up.

Teams in the “least streaky” category are pretty trustworthy when they have a lead, since they are not very likely to allow the opposition back into the game quickly. However, teams like SMU and Miami FL often struggle to overcome a deficit quickly. Look to rely more on these teams when they are favored, but be cautious about counting on them to win several games as underdogs.

Lastly, we touch on the “suspect” group, which comprises teams prone to giving up scoring runs to the opposition while struggling to counter on their own. There are a lot of bad teams on this list, but it also notably includes Villanova and UCLA, who have given up a ton of 10-0 runs against high-major opponents.

Now, what does all of this mean for your team’s chances of reaching a Final Four?

Let me first state that while the Kill Shot is a helpful tool, it can’t be used in a vacuum to determine the best teams in the country. Instead, I view it as a useful companion to other metrics, providing more nuance on the characteristics of each team.

With that said, since the 2018-19 season (which is as far back as my Kill Shot database goes), heading into the tournament, every national champion has had an Adjusted Kill Shots Per Game of at least 0.47, an Adjusted Kill Shots Conceded Per Game under 0.34, and an Adjusted Kill Shots Margin of at least 0.28.

13 teams qualify using this year’s statistics: Michigan, Arizona, Illinois, Florida, Duke, Saint Mary’s, Nebraska, St. John’s, Michigan State, Arkansas, Virginia, Vanderbilt, and Tennessee.

In general, I have found that preventing 10-0 runs is the more predictive stat of teams that advance the furthest in the tournament. While it’s great to be able to put up scoring runs in a hurry, the best teams are the ones who can hold a lead and not quickly allow an opponent back in the game.

Opponent Strength Adjustment

This is a new graphic this year that looks at how much better teams play in elite competition compared to when they play weaker teams on their schedule. The Opponent Strength Adjustment metric, which you can find on the Team Ratings page at EvanMiya.com, measures how well each team performs above or below expectations based on whether they are playing a really good opponent or a really bad opponent. Some teams perform better than usual when they face really tough opponents, but struggle to dismantle weaker opponents. This concept is sometimes referred to as “playing up/down to competition”. Other teams are the opposite: they can easily crush inferior opponents but are disappointing against higher-quality teams on their schedule. Each team has a variable that measures where they are on this spectrum of playing up or down to their competition.

This graph compares each tournament team’s predictive team efficiency rating on the horizontal axis with their Opponent Strength Adjustment on the vertical Axis. Teams way above the middle line save their best performances for really good teams, but are more lackluster as comfortable favorites. Teams way below the middle line often beat up on inferior opponents, but have not consistently performed as well as expected against the elite teams on their schedule.

All of the 1-seeds are at or above the middle line, which means that they all play well in big-time matchups, which is a characteristic consistent with most of the title-winners in recent years going into the tournament.

There’s a group of dangerous teams in the “High Ceiling Low Floor” group: Wisconsin, Michigan State, Nebraska, Texas Tech, and TCU. These teams have all been pretty awesome in big-time games but have often struggled against the weaker teams on their schedules. They are potentially more upset-prone in the first round or two, but if they can make it past those games, they could cause real problems as an underdog against an elite team.

Several Cinderella candidates have played really well against the best teams on their schedule: Hofstra, Miami OH, Troy, and Kennesaw State.

On the flip side, there are five teams in the “High Floor Low Ceiling” group: Alabama, Kentucky, Louisville, NC State, and Villanova. These teams, Louisville and NC State in particular, have really struggled to put up convincing performances against great teams. They are probably more upset-proof than other teams on their seed line, but haven’t shown the capability of proving themselves against the top-tier teams in the sport.

Consistent or Volatile

Last big chart, but one, I find very interesting as well: tracking game-to-game consistency in teams. The “Game-to-Game Volatility” metric measures how much a team’s performance level fluctuates between games. Teams higher on this metric are more volatile, while teams in the lower area of the graph are more consistent in their performance.

Arizona and Duke are among the more consistent teams in the nation, which is probably a good thing for their title hopes, given they are way more talented than pretty much everyone else.

The most volatile teams in this tournament are Utah State, NC State, Saint Louis, Texas Tech, Georgia, and Wisconsin. They all have pretty big swings between their best performances and their worst. As an underdog, this is probably a good thing as it gives them a better chance of causing an upset if they are at their best. However, it does make them less reliable over multiple games to sustain a high performance level.

Matchup Previews

To dig more into the individual matchups, the Matchup Preview Tool at EvanMiya.com lets you analyze any hypothetical matchup in the tournament. It gives you score predictions, injury adjustments, other matchup-specific adjustments, a team comparison, a roster breakdown, and a daily updated injury report.

This preview shows the potential top-10 matchup in the 2nd round between Iowa State and Kentucky:

Tournament Probabilities By Region

The tables below show each team’s chances of advancing to each round of the tournament, which you can also find at EvanMiya.com, updated daily as the tourney progresses.

Mega Bracket Breakdown Episode

Tomorrow morning, I’ll be releasing a massive region-by-region breakdown of the NCAA tournament bracket, along with extended analysis on many of these graphics here. Be sure to subscribe on podcast platforms or on Youtube!

Though coaches can optimistically believe that they can get a tune out of any roster they have, team success is highly correlated with roster talent. Only the very best coaches can significantly overcome a lack of individual quality on their team.

A large part of each team’s preseason projection on EvanMiya.com comes from a Roster Talent grade I assign to each team. This grade adds up the talent on a roster by taking a weighted sum of each player's projected rating, using Bayesian Performance Rating, and accounting for which players are likely to play more minutes than others.

The graph below compares each team’s preseason roster grade with its current team efficiency rating on EvanMiya.com for the 2025-26 season. This visual shows how much teams can perform significantly above or below the expectations for their roster.

As you can probably see, there is a pretty tight correlation between preseason roster grade and how the season has panned out for most teams. In fact, around 80% of the “variation” in current team rating can be explained by the preseason roster grade. In other words, a team’s season is 80% defined by how good the roster is before the year even starts, with the remaining 20% being everything that happens after that. There are so many factors that can explain why a team’s season plays out the way it does: coaching ability, in-season player development, scheme changes, injuries, locker room chemistry, sickness, etc. Yet roster quality is by far the most important variable, and only the best coaches can consistently outperform their talent.

Outperforming Roster Talent

To put some more concrete numbers to this, I will define a team that has “outperformed expectations” as one in the top 15% nationally, performing better than expected given its roster talent. Similarly, a team in the bottom 15% has “underperformed expectations”.

The graph below displays these bands: If a team is above the gray shaded area, it means they have significantly outperformed expectations for their roster, and if a team is below the gray area, it means they have failed to meet expectations. Every team that falls in the gray area is in the middle 70% of teams performing within a reasonable range of what was expected of them based on their Roster Talent grade.

Looking at the teams likely to make the Big Dance, there are some notable overachievers, listed in the table below. Michigan and Duke qualify because, even though their rosters were top-five to begin the year, they have turned out significantly better than expected. Other national title contenders like Arizona, Michigan State, Iowa State, and Nebraska fall on this list. Both Santa Clara and Saint Mary’s out of the WCC have moved up more than 50 spots compared to their preseason roster talent ranking.

Every coach hoped, and maybe even believed, that they could be on the “overachievers” list by the end of the season. For 85% of them, being on that list didn’t turn out to be the reality. Most teams fall within the “within expectations” band, with some falling short of even that.

This zoomed-in version of the previous graph shows every team ranked between 31st and 80th in the preseason based on their roster talent grade at EvanMiya.com. Out of this group of 50 teams, 8 of them are labeled as “overachievers”, 7 of them are labeled as “underachievers”, and the remaining 35 are within range of where they were expected to be based on their roster talent, which represents how most teams’ seasons will go.

The main takeaway here is how difficult it is for most teams to end up significantly better than their roster grades, at least according to the player projections at EvanMiya.com. In fact, over half of the teams in this group are within 15 spots of their preseason roster talent ranking.

The table below shows each team’s preseason roster ranking compared to their current team ranking for those in the “meeting expectations” category.

Making Smart Roster Decisions

The importance of obtaining a quality roster is not something unique to college basketball. In every team sport, it’s almost impossible to win a championship of any kind without being among the best in talent in that pool of teams. For college hoops, that not only matters for national championships but also for conference titles.

Given how much unavoidable roster turnover there is for most teams these days, coaching staffs must have roster-building tools that help them make more accurate player evaluations. There are a ton of good ones out there, but I’ll advocate for many of the resources at EvanMiya.com that have been invaluable to teams for years. The player evaluations that teams can access on the site are the same ones that calculate these roster talent grades mentioned today. If your goal is to have a top 50 team in the sport next year, but you only have a roster talent grade of 80th in the preseason, evidence shows that you don’t have a good chance of overcoming that roster diagnosis and achieving the success you want.

There are two main product tiers at EvanMiya.com that I recommend for roster building and player evaluations. The first is the “Premium” subscription, which is available to anyone. This tier gives you projections for every player in the transfer portal, plus dedicated player skill projections pages for every Division 1 player. You can read more about the player skill projections in this previous article.

The most powerful tool I have for roster building is a new set of features I launched last year called the Front Office Suite, which is specifically designed for coaches, GMs, and ADs who want to use player projections to actually obtain NIL value estimates for every single player in Division 1. There were a bunch of notable teams, including Michigan, who signed up right after the tool launched last April and used it to help with NIL evaluations, build intelligent player searches, assess hypothetical lineup chemistry, get estimates of their roster talent, and more. If you’re interested in getting more info on this product, email me at evanmiyainfo@gmail.com or send me a direct message.

More March Analysis

I’ve got a ton of exciting March Madness content planned, with a lot of it coming on my new weekly podcast, The EvanMiya College Basketball Show! It’s a weekly podcast with episodes every Tuesday morning. The show covers national storylines and recent trends in CBB, featuring multiple guests each week — plus a touch of analytics.

There will be a new episode out tomorrow morning (Tuesday the 3rd). For new, here is the latest show from this week:

This season, the SEC has taken a significant step back. The conference has just one team in the top 12 nationally at EvanMiya.com, has only nine teams deserving to be in the field right now according to Resume Quality, and may not land a single team on the one or two-seed line on Selection Sunday.

So why is the SEC still rated as the best conference in basketball at KenPom.com, and (previously) EvanMiya.com?

The answer makes sense numerically but still feels unsatisfying. If you take the average predictive team efficiency margin across every team in each conference, the SEC still shakes out as the strongest conference in the country. This is mainly because the middle and bottom tiers of the conference are still comfortably better than every other league. The SEC has over 80% of its teams inside the top 60 nationally at EvanMiya.com, and they have only a single team outside the top 100 (South Carolina at 106th). Conversely, the Big 12 has just 9/16 teams inside the top 60 nationally, and has five teams outside the top 75, compared to just one in the SEC. A third of the teams in the Big Ten rank outside the top 60, and it includes three very weak high-major squads this year: Maryland (111), Penn State (123), and Rutgers (137).

So, yes, the SEC still technically has the hardest group of teams to beat from top to bottom… But that isn’t REALLY what we care about, is it?

Creating A New Conference Ranking System

When we look back in mid-April and memorialize the 2026 season, each conference will be primarily remembered for a few things:

1. Did your conference win a national title?

2. How many Final Four teams did your conference have?

3. How many teams did your conference get into the tournament?

Though the SEC has the strongest group of teams from top to bottom, both the Big Ten and the Big 12 are way more likely to move the needle on a national level in March. The Big Ten has 5 of the top 11 teams at EvanMiya.com right now, and 8 teams in the top 35. The Big 12 has 6 of the top 16 teams in the country.

To better measure the top-end quality of each conference, I have created a new conference strength metric, which you can find updated daily on EvanMiya.com's “Misc” page. I’m calling it “Upper Strength” to represent the team quality in the upper regions of each conference. I’ll share the exact details of how this is calculated in the footnotes1, but essentially, teams at the top of the conference are weighted much more heavily than those at the bottom.

The table below shows the updated conference strength standings with this new metric:

The ordering of the top conferences much more accurately reflects the high-end quality in each league. The Big Ten and Big 12 are neck-and-neck for the top two spots, with the SEC slotting in a bit behind in third. The ACC and Big East form the next tier, followed by the WCC, Mountain West, and Atlantic 10.

If you equally weight all teams in each conference, the WCC actually rates as the 8th-best conference nationally. However, using the Upper Strength conference metric, it surpasses the Mountain West and Atlantic 10, largely because it has three at-large-quality teams: Gonzaga (12th), Saint Mary’s (40th), and Santa Clara (43rd). When we talk about how good the WCC is this year, no one really cares that Portland and Pepperdine are outside the top 250 teams nationally.

Diving Deeper

I will go further into the implications of this new conference ranking system next Tuesday on the upcoming episode of my new show, The EvanMiya College Basketball Show! It’s a weekly podcast with episodes every Tuesday morning. The show covers national storylines and recent trends in CBB, featuring multiple guests each week — plus a touch of analytics.

Here is the latest show from this week, featuring Kevin Sweeney from Sports Illustrated:

1

The exact formula for assigning a weight to each team in the conference is as follows: Give each team a numerical ranking R from 1 (best) to N (worst) in the conference. Determine each team’s weight by the following: weight = [1 - ((R - 1) / (N - 1))] ^ 2.

This exponential weighting system assigns significantly more weight to teams ranked higher in the conference, with diminishing weight for teams in the middle and lower tiers. The table below shows how this works for the SEC conference. A team ranked in the dead middle of a conference will have about 25% of the weight of the best team in the conference.

Many programs followed the herd in a sudden rush to pan for gold that seemed to appear out of nowhere. The question is: did many of them actually strike it?

In this article, I’m going to present the stats on production from international recruits to assess their effectiveness.

For the sake of this discussion, I am focusing on one group: international players with prior professional basketball experience who are in their first season of Division 1 college basketball. This does not account for international transfers who have already played a season in college (Matas Vokietaitis transferred from FAU to Texas), nor am I considering players who are not native to the US but were ranked by recruiting services like 247Sports well before the summer of 2025 (Dame Sarr is an example). I’m specifically interested in international recruits who received offers to play in college during the past offseason. My list of international players will not be exact, but it is very representative of the overall crop we’ve seen enter college basketball this season.

Most Productive First-Years

I think the easiest way to evaluate the effectiveness of these international imports is to compare their production to the rest of the “traditional” freshmen class. For much of this study, I’m using Box BPR as the measuring stick of productivity. For those who are unfamiliar with Box BPR, it’s a college version of Box Plus-Minus that looks at your individual box score statistics and quantifies how valuable a player you are based on your stats alone.

The table below ranks the top 30 first-year players in college basketball according to Box BPR. The ten most productive first-years are all traditional freshmen, but seven of them were international recruits this offseason.

Most of the standout international guys are on this list, including Hannes Steinbach, David Mirkovic, Mario Saint-Supery, and Thijs De Ridder. Other notable names aren’t that far behind, such as Johann Grünloh, Ivan Kharchenkov, and Sananda Fru.

Comparing Freshmen and Internationals

If we compare the overall validity of recruiting international freshmen as opposed to getting traditional freshmen, how do these two groups stack up?

If you take the top 50 most productive freshmen this year, the average Box BPR of that group is +5.1. For context, the entire SEC conference has about 30 players with a Box BPR above +5. That’s a very good mark for a group of 50 freshmen. By comparison, the top 25 most productive international first-years have an average Box BPR of +3.2. You have to whittle the international list down to just the 11 most productive players to match the average level of the top 50 traditional freshmen. Basically, only the top 11 international recruits have been as productive as the top 50 traditional freshmen. That list of 11 internationals is, in order: Hannes Steinbach (Washington), David Mirkovic (Illinois), Mario Saint-Supery (Gonzaga), Thijs De Ridder (Virginia), Ruben Dominguez (Texas A&M), Massamba Diop (Arizona State), Roman Domon (Murray State), Sananda Fru (Louisville), Johann Grünloh (Virginia), Ivan Kharchenkov (Arizona), and Stefan Vaaks (Providence).

I’ll take this a step further. The dream for many of the high-major teams that pulled the trigger on an international pro this summer was to find themselves a hyper-productive player out of thin air. For most teams, if that was their goal, they have been disappointed.

If you take the top three most productive players on each power-conference team this season, according to Box BPR, there are 25 traditional freshmen on that list. This means that 25 traditional freshmen are a major driving force behind their team’s success this season. The list of those freshmen is below:

In comparison, there are just nine international first-years who rank in the top three on their own team in production among the big five conferences. By my count, that means 31 other internationals are playing for high majors that have not moved the needle on a national level in a significant way.

I bring all this up to offer a word of caution: When we get to the bright lights of March Madness, many color commentators and studio analysts will point to the incredible talent on display that has come from an influx of international professional talent, and rightfully so. But for every Thijs De Ridder that we praise, several other international guys have faded into anonymity that people will quickly forget about. The hit rate on these players is not anywhere close to automatic.

If you look at the top 50 most hyped international players coming into the college basketball season, according to different recruiting services and analysts, more than half of them this year have had pretty lackluster statistical output. In fact, 28 out of the 50 have a Box BPR that currently sits outside the top 1500 players in the nation. Of course, some of these players still play important roles for their teams, but they have not been lighting it up like some may have hoped.

Here is the full list of international first-years who got some offseason buzz that are outside the top 1500 in Box BPR, in alphabetical order: Achille Lonati (Saint Bonaventure), Andrej Kostic (Kansas State), Andrija Jelavic (Kentucky), Dominykas Pleta (Iowa State), Efe Demirel (Grand Canyon), Ege Demir (Oregon), Elias Rapieque (Kansas State), Evan Boisdur (Grand Canyon), Harun Zrno (Rutgers), Ian Platteeuw (Davidson), Ilia Ermakov (Saint Bonaventure), Jeff Nwankwo (Oklahoma), Jordi Rodriguez (Cincinnati), Lefteris Mantzoukas (Oklahoma State), Mantas Juzenas (Saint Mary’s), Mathieu Grujicic (Ohio State), Mihailo Petrovic (Illinois), Musa Sagnia (NC State), Ognjen Stankovic (Akron), Omer Mayer (Purdue), Paul Mbiya (Kansas), Ron Zipper (LSU), Sidi Gueye (Arizona), Tama Isaac (UC Irvine), Tibor Mirtic (Penn State), Timotej Malovec (Miami (Fla.)), Vangelis Zougris (Louisville), Yohann Sissoko (Florida Atlantic).

Are International Recruits Meeting Expectations?

I’ve left full Bayesian Performance Rating out of the conversation until now for a reason: I had to make projections for all of these international recruits before the season started. Based on recruiting ratings, statistical data, and predictions from many sources about these players, I made a judgment call on how good I thought all of these international first-years would be. In general, I was VERY conservative with these projections. I was criticized by several people for not being bullish enough on how impactful this crop of internationals would be.

In the end, these projections were too pessimistic on some of the players, but a majority of the international first-years have been even worse than where I slotted them in the preseason. The table below shows the current Bayesian Performance Rating of all the international recruits this year, from best to worst. In addition, you can compare their current rating to what their preseason projection was, and see if their ranking among all freshmen has moved up or down since the year started.

Of the 53 international first-years on this list, only 16 of them (30%) have improved their ranking among all freshmen since the year began. The vast majority of them have been worse than I projected. Players like Mario Saint-Supery, David Mirkovic, Johann Grünloh, Thijs De Ridder, and Hannes Steinbach have been outstanding. The same can’t be said about most of the rest.

There is a big difference in quality between the top international first-years and many of the rest. Teams like Virginia, who scouted well and put the right resources forward to get the quality players, have been well rewarded. Other teams that hoped for similar results but didn’t evaluate overseas talent well have been left wanting more. This analysis reinforces the importance of good scouting and evaluation.

If you want more analysis, I have a new platform this year: The EvanMiya College Basketball Show! It’s a weekly podcast with episodes every Tuesday morning. The show covers national storylines and recent trends in CBB, featuring multiple guests each week — plus a touch of analytics.

Here is the latest show from this week, with Jeff Borzello from ESPN:

The EvanMiya College Basketball Show will drop new episodes every Tuesday morning. The show covers national storylines and recent trends in CBB, featuring multiple guests each week — plus a touch of analytics.

The first episode is live, and can be found on all the usual podcast platforms, as well as on YouTube.

On the debut episode:

• John Fanta joins to talk Nebraska’s run, Michigan’s loss, UConn/Houston, and title favorites

• Michigan Wolverines deep-dive with Greg Waddell

• Predicting 1 seeds

• Efficiency tiers, and other analytics nuggets

Subscribe to the podcast here!

And don’t worry, there will be more regular written content here on the blog over the next couple months!

I finally got around to producing preseason lineup efficiency projections and grades for a bunch of possible lineups that every team could use regularly during the season. You can find the entire list of lineups, sortable by predicted efficiency or individual statistical projections, here on the Projections / Grades section of the Lineup Metrics. The table below shows the top 50 lineups nationally, ranked by predicted efficiency margin this season:

Each lineup can also be clicked individually, bringing up a page that shows that lineup’s projections and skill grades in greater detail, compared to the rest of Division 1. Additionally, you can select just a single team to see all of their possible lineup combinations and how they are rated.

To gain some interesting talking points from these projections, I’ve picked the best lineup in every statistical category, which I will reveal below.

Best Overall: Florida

Florida’s possible starting five has the best predicted efficiency margin (including offense and defense together) of any possible lineup in the country. In the model, this lineup gets slightly penalized for how tall it is (Thomas Haugh would be playing at the 3 as a 6’9” forward), and it still ranks as the best in the country. For what it’s worth, any combination of Condon, Handlogten, and Chinyelu at the 4 and 5 spots still gives Florida the best lineup projection in the country.

Best Offensive: Purdue

Purdue has the #1 offense in the preseason at EvanMiya.com, has the best player in the country in Braden Smith, and boasts the best offensive lineup. Though this unit, featuring Liam Murphy instead of CJ Cox, is not Purdue’s best overall lineup in my projections, it has the potential to be lethal on the offensive end.

Best Defensive: Kentucky

Kentucky will start the season with an incredibly imposing defense, and it will only get scarier once Jayden Quaintance becomes healthy. The Wildcats will have a number of lineups among the nation’s best on defense, including their top defensive lineup in these projections with Lowe, Oweh, Dioubate, Quaintance, and Garrison. This same lineup with Kam Williams replacing Garrison also has a better defensive projection than any other team can match.

Best Three-Point Shooting: Purdue, Louisville

I’m including Purdue and Louisville here since both deserve a shoutout for their prolific three-point shooting. Purdue’s possible starting lineup is predicted to shoot 39% from three, while Louisville clocks in at 38%.

Best Two-Point Shooting: Creighton

This Creighton unit is predicted to shoot almost 60% from inside the arc.

Best Free Throw Shooting: Louisville

Best Assist Rate: Florida, High Point

Florida’s possible starting five, mentioned previously, has the highest predicted assist rate in the country, but I wanted to mention High Point’s lineup, which ranks 2nd nationally in this category.

Best Turnover Rate: Sacred Heart, Creighton

Sacred Heart and Creighton have the two lineups in the country with the lowest predicted turnover rate.

Best Offensive Rebounding: Baylor

Best Defensive Rebounding: Iowa

Best Steal Rate: Iowa State

Best Block Rate: San Diego State

Most Experience: Oklahoma

Tallest: Illinois, Duke

I don’t know if Illinois would play this lineup, but it sure would be fun: Kylan Boswell 6’2”, Ty Rodgers 6’6”, Andrej Stojakovic 6’7”, Tomislav Ivisic 7’1”, and Zvonimir Ivisic 7’2”.

This Duke lineup would be 2nd in average height behind only Illinois: Cayden Boozer 6’4”, Dame Sarr 6’8”, Maliq Brown 6’9”, Cameron Boozer 6’9”, and Patrick Ngongba 6’11”.

As always, if you want to support my work, sign up for a subscription at EvanMiya.com, it’s very appreciated. I’m so grateful I get to do this for a living.

Instead of focusing solely on where each team is placed in my preseason team rankings at EvanMiya.com, a more interesting angle is to use the uncertainty estimates attached to each team’s preseason projection to determine their ceiling and floor outcomes. Using 10,000 season simulations, I have calculated various possibilities for how the season will unfold.

Some teams have more uncertainty attached to their preseason projection than others. These teams usually have common themes: a new head coach, more reliance on freshmen, or more roster turnover. Teams like Auburn, Baylor, and Duke have more variance, while the model is more confident in the possible outcomes for teams like Purdue, Iowa State, and Florida.

To present this data on a national level, I’m breaking it down by conference and presenting the median (most likely), ceiling (95th percentile), and floor (5th percentile) outcomes for each team. This means there is a 90% chance a given team will have a national ranking within their ceiling and floor outcomes at the end of the year.

First, though, I want to briefly mention the teams with the highest and lowest levels of preseason uncertainty according to my model. The table below shows the teams with the most preseason uncertainty (high-ceiling, low-floor) inside the preseason top 50 at EvanMiya.com:

The table below shows the teams with the least preseason uncertainty (low-ceiling, high-floor) inside the preseason top 50 at EvanMiya.com:

Only six teams have a ceiling finish of #1 in the country: Houston, Purdue, Duke, Florida, UConn, and Michigan. While another team can be the best-rated team at the end of the year according to the analytics, it’s much more unlikely than for these seven. In fact, only three times in the last 15 years has a team finished #1 at EvanMiya.com without having a preseason ceiling finish of #1 (UConn 2022-23, Villanova 2017-18, Gonzaga 2016-17). Let me also be clear that a team’s ceiling finish does not refer to their chances of winning a national title. In college basketball, the best team in the sport often does not win the ultimate trophy at the end of the season.

Here are the teams presented by conference, from best conference to worst, according to the preseason conference strength rankings at EvanMiya.com:

1. SEC

2. Big Ten

3. Big 12

4. ACC

5. Big East

6. Atlantic-10

7. Mountain West

8. WCC

9. American

10. Missouri Valley

11. Conference USA

12. Ivy

13. MAC

14. CAA

15. Big West

16. SoCon

17. Sun Belt

18. Big South

19. Big Sky

20. WAC

21. Southland

22. Summit

23. MAAC

24. Horizon

25. Patriot

26. Atlantic Sun

27. Ohio Valley

28. America East

29. SWAC

30. NEC

31. MEAC

This offseason, I made some significant upgrades to the Bayesian Performance model that addressed some lingering biases and led to sharper model accuracy. Rather than just mentioning these changes quickly, I wanted to provide users with a complete technical examination of the model so that I can properly convince readers that Bayesian Performance Rating is the best all-in-one player metric publicly available in college basketball.

If you want to skip the explanation altogether, jump to the tables at the end or go view the full datasets on the Player Ratings page at EvanMiya.com.

The Goal of BPR

In a nutshell, Bayesian Performance Rating reflects the offensive and defensive value that a player brings to his team when he is on the court, on a per-possession basis. This rating incorporates aspects such as individual efficiency statistics, on-court impact on team performance down to the play-by-play level, previous career performance, high school recruiting profiles, adjustments for opponent players and teammates, and more. All of this results in one offensive and defensive number, which can be added together to get a player’s overall Bayesian Performance Rating.

Here’s the technical interpretation of BPR:

Bayesian Performance Rating represents how many points per 100 possessions better a player’s team is expected to be than its opponent if that player were on the court with nine other average Division I players. For example, if a player has a BPR of +5.0, this means that if he were playing with average D1 teammates, against average D1 players, his team would be expected to outscore the opposition by 5 points in a 100 possession game.

Typically, the best player in the country each year finishes with a BPR above +10.0, a top 50 player in the country would have a BPR of about +7.0, an average high-major starter would have a BPR of about +4.0, and an average D1 player would have a rating of precisely 0.0.

Part of what makes BPR unique is that it is a predictive stat: BPR doesn’t just summarize how good a player has been in a given season, it actually predicts how productive that player will be going forward. This makes BPR way more useful during a season because you don’t have to wait till February for it to become accurate. Players even have BPR predictions in the preseason, which will start to go up or down as the model updates with current season performance data.

Ultimately, the goal of Bayesian Performance Rating is to get player values that lead to the best future predictions of player and team performance. Every part of the BPR formula has been optimized to achieve the best out-of-sample predictions, employing various cross-validation methods. When a player has a BPR of +6.0, we genuinely believe that he will add about 6 points of value per 100 possessions to his team compared to an average player, based on our extensive testing. There are other commonly referenced player efficiency formulas, such as PER1 or Offensive Rating2, which are human attempts to weigh each part of a player’s box score to measure their overall contribution towards winning. While convenient, these model weights are not determined by rigorous statistical testing and are therefore less reliable for future predictions of player performance.

Here is the formula used for predicting the expected number of points scored in a single offensive possession:

In this equation, i is the intercept representing the average points per 100 possessions in CBB, approximately 104, with a home team advantage adjustment if the game is not played on a neutral court. Each B term represents the coefficient, or BPR rating, for an individual player on offense or defense. If you add up the Offensive BPRs of the five players on offense, and you subtract the Defensive BPRs of the players’ defense, you get the expected points for that possession.

Many model features make BPR particularly effective, which we will unpack in greater detail in a moment:

• The model accounts for individual player statistics and advanced box score metrics.

• The model measures impact on team performance through analyzing play-by-play outcomes.

• The model adjusts for the strength of all other players on the floor for every possession played.

• The model is trained on years of Division 1 college data to specifically measure what’s important in the college game.

• BPR is a predictive stat, estimating how valuable a player will be going forward.

• The model comes with preseason projections for each player, using player historical information to make it more accurate. BPR adjusts up or down as the season progresses, making it useful the entire year.

• BPR is bayesian in nature, which allows for preseason priors and measurements of uncertainty around each player’s value estimate.

Model Comprehensive Breakdown

I’m going to explain how each component of Bayesian Performance Rating works. To accommodate readers with varying levels of interest, I will break each part into several sections:

🧭 High-level Overview

🔎 Technical Details

💡 Why It Matters

For those who just want to understand the basics, I’d recommend skipping the 🔎 Technical Details sections.

Model Structure

🧭 High-level Overview

The Bayesian Performance Rating model consists of several individual models that are carefully pieced together. These include my own versions of a regularized adjusted-plus-minus model, a box plus-minus model, and a preseason projection model, all wrapped within a bayesian predictive framework. All three of these model classes are commonly used within basketball for player evaluation, but each has its own shortcomings when deployed in isolation. BPR uses all three model types together to form a superior final version.

🔎 Technical Details

While I will expand on each model type in future sections, I want to briefly describe the individual models and mention their big-picture strengths and weaknesses.

Regularized Adjusted Plus-Minus models3 are a great way to measure a player’s overall on-court impact on offense and defense. It’s basically plus-minus, but adjusting for many of the contextual variables that are lost in basic plus-minus. These models provide valuable information about the value a player brings while on the court, without being biased towards certain player archetypes, like volume scorers or stat-stuffers. A player’s rating considers the scoring outcome of every possession played, adjusting for the strength of both teammates and opposition players faced on each possession.

Without other statistical input, the RAPM model outputs face a variety of issues. Players on very good or very bad teams are often clustered together, since the model can struggle to differentiate between them. Individual games can have a significant impact on the final player coefficients in ways that, without more context added, lead to misleading results. Sometimes players can get “lucky” and end up with a good rating, just because they happened to be on the court during one good team performance and not on the floor during a bad one. A pretty large sample size is needed to get sensible results from the model, so season-level RAPM coefficients aren’t useful till a good portion of the way through the year.

Box Plus-Minus models are an attempt to use box score statistics to create a much more stable version of RAPM. Essentially, a linear regression is performed to determine which box score statistics are most predictive of a good RAPM coefficient. Once these weights are determined, a player’s BPM can be calculated purely using advanced box score stats. BPM models are popular because they are much easier to interpret and understand, and their results are more predictable since they typically align closely with a player’s statistical output.

While BPM models are very convenient, they still leave a lot of valuable player information on the table that’s not captured in the box score. Some players don’t stuff the stat sheet but excel in many other intangible areas, leading to winning outcomes for their team regularly. Other players can score in spades and post ridiculous box scores, but may not actually be “winning” players. Like RAPM, Box Plus-Minus is also not stable with small sample sizes. Typically, BPM leaderboards need to be filtered to show players with a specific possession count to be more trusted. Because of this, BPM is also not very useful early in the season. One other important note: almost all public BPM values come from a formula trained on years of NBA data. While copying this formula and using it for college players is helpful, specific player skills are more valued in the college game, so using an NBA-trained formula leaves a little to be desired (more on that later).

Preseason projection models are not just specific to players, as they are used all across the sports analytics space. Readers will likely be very familiar with college basketball team projections in the preseason, like at EvanMiya.com, KenPom.com, and BartTorvik.com. Similar to how these team projections are formed, a player’s rating in one of these aforementioned player value models can also be predicted using previous season historical data for each player. Naturally, a preseason projection for a player is helpful at the beginning of a season. Still, it should not be viewed as the most accurate representation of a player’s value once we get deeper into the year.

💡 Why It Matters

Bayesian Performance Rating combines three different model types to give the most accurate assessment possible of player value in college basketball. While each of these individual models has significant shortcomings on its own, using their strengths together leads to much better predictions of future player performance, as I have determined from my years of rigorous statistical modeling and testing.

The following several sections will explain how each of these model components works together in greater detail.

Step 1: Multi-Year RAPM Model

🧭 High-level Overview

At its core, Bayesian Performance Rating is a Regularized Adjusted Plus-Minus model. Every Division 1 player is represented by an offensive and defensive coefficient, which measures how much more valuable they are than an average player, per 100 possessions. A coefficient above zero means a player is better than average, and a coefficient below zero means he is worse than average. The model takes an entire season’s worth of play-by-play data across all Division 1 games and adjusts each player’s offensive and defensive ratings until the set of all ratings best explains all the play-level results that have occurred.

🔎 Technical Details

As mentioned in a previous section, the regression equation used in the RAPM model takes the following form:

In a typical linear regression, the possession-by-possession data would be fed into a linear regression engine that spits out the coefficients that make the most sense. However, using a linear regression without “shrinkage” leads to nonsensical coefficients, especially for players who play very few minutes. The standard technique used is to apply some sort of regression to the mean. The simplest way is to use ridge or lasso shrinkage methods to make it difficult for players with low possession counts to be rated way above or below the league mean.

From the inception of Bayesian Performance Rating, I have chosen to use bayesian methods instead, hence the name. Using a bayesian linear regression offers two main advantages. The first is that we can be way more specific about player-level prior distributions. Instead of fixing every player’s prior mean to zero, or some other pre-determined constant, we can choose whatever prior mean and standard deviation we want for each player. As I will explain in Steps 3 and 4, it’s highly beneficial to make the RAPM model more informed by giving each player a starting value that’s a best guess of their coefficient, both in terms of specifying the mean and the amount of uncertainty we have in each player’s prior mean.

When building BPR from scratch, the first step is to create several multi-year RAPM model datasets, each using four years of consecutive data. For example, I used all play-by-play data from the 2021-22 season through the 2024-25 season, treating it as one big season and making a convenient (and wrong) assumption that a player’s value is constant across all seasons played in that multi-year sample. For these models, all players are given the same prior mean and prior variance. While none of these coefficients are actually used for anything public, they do help determine the best amount of prior variance to give to a typical player, which I determined through cross-validation methods. Tuning this parameter through optimization leads to the lowest out-of-sample possession prediction error. Additionally, the multiple years of data led to more stable and trustworthy player coefficients compared to if I had just used single seasons to build the RAPM dataset (which I had done previously).

💡 Why It Matters

Using an advanced plus-minus model on its own only goes some of the way toward building an accurate player metric that predicts future performance. There are some other commonly cited RAPM models used for college basketball, but without a lot more context, I would not feel comfortable trusting these model outputs to guide my own assessments of players. Bayesian Performance Rating is structured like an adjusted plus-minus model, but the model is fed way more information about each player through the form of prior distributions, helping the model make much more intelligent final conclusions.

Step 2: College Box Plus-Minus Model

🧭 High-level Overview

Using the multi-year RAPM datasets from Step 1, I have formed my own college-specific version of a Box Plus-Minus model, called Box BPR. BPM models for player evaluation are powerful because of their intuitive interpretability. A player’s final value is calculated by adding up different statistical box score outputs, each multiplied by some weight that determines how valuable each skill is towards winning. Each rating represents the value that a player adds per 100 possessions to his team. Categories like three-point shooting, assists, and blocks are emphasized, while adverse outcomes like turnovers and missed free throws are penalized. Each player gets an Offensive and Defensive Box BPR based on how valuable their statistical output is, according to the model. These values are used in conjunction with the RAPM model to form more accurate player ratings in Step 3.

🔎 Technical Details

A box plus-minus formula is created by determining how much weight each statistical category should get such that it best predicts overall player performance, team performance, and ultimately, winning. The way this is typically done is by taking offensive and defensive player impact values, calculated through some model that is not biased by box scores whatsoever, and then using the season-level player box statistics to predict those player impact coefficients. The most well-known Box Plus-Minus formula, derived by Daniel Myers4, is trained on several multi-year NBA RAPM datasets. Instead of copying this formula and assuming it is accurate at the college level, I created my own version, called Box BPR, trained on my own multi-year RAPM data from the previous step. All box-stat predictor variables are given to the model on a per-possession basis, such as assists, turnovers, offensive and defensive rebounds, twos, threes, free throws, etc. For example, a player’s rebounding ability would be represented as rebounds per 100 possessions.

Some important executive decisions have to be made about what kinds of predictor variables to give to the model. For example, the average opponent rating faced by each player is used to properly account for competition level. One primary choice I made was to include some measurements of playing time and usage in the model to reward players who play a larger percentage of their team’s minutes and carry bigger offensive loads. If the Box BPR model were only being used in isolation, I might consider using just the per-possession stats without any considerations for playing time or usage. However, since the Box BPR coefficients are used to form each player’s prior mean in the larger RAPM model, giving credit for playing load does make the final BPR model outputs more realistic. Additionally, I do not shift every team’s Box BPR values up or down to add up to the total team efficiency ratings, as is commonly practiced. Doing this skews the final BPR output in undesirable ways.

💡 Why It Matters

While other public Box Plus-Minus models are helpful, the version at EvanMiya.com, Box BPR, is more accurate in college basketball because it is trained on Division 1 data to determine what skillsets are the most valuable at this level. Additionally, since the final BPR model combines the Box BPR values with an adjusted plus-minus model, the box score statistics are properly weighted without relying on them to tell the whole picture about player performance (more on those weights in the next section).

Step 3: Forming a Prior-Informed RAPM Model

🧭 High-level Overview

This is the step where BPR really starts to take shape. Using a bayesian linear regression model allows us to create player-specific prior means, which essentially behave like starting points for the RAPM model. In this case, I use each player’s season-level Box BPR values as a starting point for their offensive and defensive coefficient. For example, suppose a player has an Offensive Box BPR of +3.5 points per 100 possessions. In that case, the model will start by setting the final player's offensive value at +3.5 and then adjust that value up or down based on how that player actually impacted the team's offensive efficiency while on the court. This really helps the model immediately identify the good and bad players, rather than naively giving everyone equal footing to start.

🔎 Technical Details

In our prior-informed RAPM model, we need to get each player’s prior distribution, which will be represented by a normal distribution with a prior mean and prior standard deviation. The prior mean for each coefficient is simply each player’s respective Box BPR rating on offense or defense for the season. The trickier part is figuring out the proper prior standard deviation for each player. I’m not going to reveal all the secret ingredients here, but essentially, the prior standard deviation is determined based on how much of the variation in the RAPM model coefficients can be explained by each player’s Box BPR values. For offense, the Box BPR explains about 50% of the variation in the RAPM coefficients, and for defense, it is about 30%. We can directly use these R-squared values to figure out the exact magnitude of the player prior standard deviations used in the bayesian model.

To make this even more accurate, I actually tested a whole grid of possible global offensive and defensive prior standard deviations to see which ones led to the best out-of-sample possession outcome predictions. I did this partly to see if giving even more weight to the Box BPR priors makes the model more accurate. Turns out, more reliance on these box score priors was useful! Through cross-validation, the best performing offensive prior standard deviation was one equivalent to an R-squared of 66%, and the best defensive prior standard deviation was one equivalent to an R-squared of 59%. What does that mean? When you look at the final BPR ratings, more than half of the weight is given to the box score statistics for most players, with the remaining weight dedicated to the adjusted plus-minus model, which quantifies impact beyond the box score.

Once these prior distributions are determined, the full prior-informed RAPM player ratings can be calculated, leading to full season-level BPR values that incorporate box score stats and adjusted plus-minus modeling.

💡 Why It Matters

Bayesian Performance Rating combines two extremely useful types of data: box score statistics and advanced on-court team impact metrics. Using these in tandem leads to a much more accurate representation of player value, especially when it comes to predicting future impact. The amount of strength given to each of these components is determined scientifically using statistical methods that figure out the proper weighting.

Step 4: Making Preseason Priors

🧭 High-level Overview

The final modeling step is to create preseason projections for each player so that Bayesian Performance Rating is useful from the very beginning of the season. These projections, similar to how team preseason projections are calculated, take a bunch of variables specific to each player to predict that individual’s value for the upcoming season. Variables used include player BPR and Box BPR in previous seasons, high school recruiting profiles, expected developmental growth by class, transfer status, and more. Once these projections are formed, they are also included in the overall BPR model as the season progresses, serving as starting points for each player before any games are played. As a new season unfolds, the BPR values will begin to go up or down based on whether players are performing better or worse than their initial preseason projection. By the end of the year, these preseason projections still carry some weight toward the final BPR ratings, but are far outweighed by the current season box score and team impact metrics used in the model.

🔎 Technical Details

Similar to Box BPR, the goal of the preseason projections is to have a preseason prior distribution for each player. Accurately estimating the prior means is crucial. These starting values have a big effect on BPR outputs early in the season, and they are meant to be useful right away to coaches and fans analyzing player value. There is another equally important step: estimating preseason uncertainty for each player’s prior. Having a unique measurement for preseason uncertainty (prior variance) for each player makes sense. Some players who have been in college for a while are known commodities. They aren’t likely to be a lot better or worse than we expect. Other players are massive wild cards. Highly touted freshmen are erratic, players stepping into bigger roles can be hit or miss, and those who previously missed an entire season may be hard to quantify. Using bayesian modeling techniques for these projections is incredibly valuable because it allows us to obtain a probability distribution for each player that represents all the reasonable outcomes we can expect for how their season will unfold.

I won’t go into much greater detail on how these projection models are built, but I will mention the main factors that contribute to each player’s prior:

• Stats and metrics from previous college seasons are used, including BPR, Box BPR, playing time, position, offensive role, and player class. This also accounts for the competition level faced in previous years.

• High school recruiting ratings are used, especially for freshmen and younger players, to determine each player's unrealized potential. For example, if two freshmen had similar first-year performances, a 5-star recruit out of high school will have a better sophomore year preseason projection than a 3-star recruit will. Freshmen (and international recruits) projections are constructed by looking at how previous players with similar recruiting profiles have fared in their first year at the D1 level.

• Transfer status is also taken into account, which is new to the model this year. I now have two different preseason models, one for returning players and one for transfers. I have found that some metrics are more accurate in predicting player performance for returning players than they are for incoming transfers. For example, advanced On-Off splits are pretty useful for predicting how a player returning to his same team will fare in the following year, but are less helpful for players who change teams and contexts. In general, the transfer model leans more heavily on the box score statistics and less on the other on-court impact metrics.

Once these preseason priors are calculated, I use a bayesian hierarchical model structure to combine the preseason prior with the Box BPR, forming a new, final prior that informs the RAPM model. As the season progresses, the preseason projection portion of the prior will become less influential on the final model outputs. Notably, though, I don’t ever completely erase the preseason data from the model. Since the entire goal of Bayesian Performance Rating is to have ratings that predict future player performance, keeping the preseason priors in, at least to some degree, does really help make better estimates of player value during postseason play. As a general rule of thumb, the preseason priors will carry about 15% of the overall model weight by the end of the season, but this varies greatly player-by-player.

💡 Why It Matters

Having these preseason BPR projections adds significant utility to the model. Not only can we obtain solid estimates of player value before the season even starts, but having these priors influence the model during the season also helps stabilize the model. BPR doesn’t overreact to early performances in the season but uses all available data on each player, including past performances and high school recruiting information, to form the best possible evaluations of players in the present.

Final Numbers

Enough technical details, let’s look at some numbers! The table below shows the top 25 players in BPR at the end of the 2024-25 season. In the top four, Flagg and Broome were clearly the best two players all season, and Braden Smith and Walter Clayton were consensus first-team All-Americans.

You can find the ratings for every other season on the Player Ratings page at EvanMiya.com, but I did want to take a look at the top 30 players all-time in BPR, going all the way back to 2009-10:

Zach Edey’s historic season in 2024 tops the list, followed by what Cooper Flagg just did last year in perhaps the most dominant season for a freshman ever. Zion Williamson’s lone season also makes the top five. Michigan State has four players in the top 30 all-time, the most of any team.

2025 Model Updates

For those who reference BPR regularly, I wanted to provide a brief list of the significant model updates for this upcoming season, although some have already been mentioned.

• All data now goes back to the 2009-10 season, which you can find on the Player Ratings page.

• I switched to using a 4-year RAPM college dataset, using several 4-year stints, to establish more accurate model tuning parameters. This led to updated shrinkage parameters, Box BPR stat weights, and overall weight given to box statistics vs on-court play-by-play data. The new model leans a little more heavily on the box score statistics, and the magnitude of the BPR values in a given year is also slightly higher than before.

• The model more properly accounts for player minutes and usage and is not as biased against players who carry a big load like before. In the previous iteration, players who were on the court for a significant portion of their team’s possessions sometimes struggled to get the credit they deserved. For example, high usage guards like Devin Carter (2023-24), Walter Clayton (2024-25), and Bennett Stirtz (2024-25) have received significant boosts in the new model version.

• Excellent players outside the high major ranks have also received significant bumps in some cases, such as Yaxel Lendeborg, DaRon Holmes II, and Jonathan Mogbo.

• The preseason player projections model now includes two separate models based on whether a player is transferring or returning to his old team. Often, the variables that predict player success vary based on whether a player is returning to the same team and system or not. For example, advanced On-Off metrics are pretty useful for assessing player performance, but they are significantly less reliable for transfers.

Due to these updates, all previous season pages have been refreshed to reflect the new end-of-season BPR and Box BPR values using the latest model.

Here’s a preview of the top 10 teams and players on the site:

Quick Methodology Overview

For those who aren’t familiar, here’s an overview of how the team preseason projections are formulated:

• Projected roster strength - This is by far the most important one. Each player on the roster has a projected Bayesian Performance Rating for the season, which estimates how impactful a player will be on a per-possession basis during the season. There is also an associated minutes prediction model that gives a rough estimate of what a team’s rotation will look like. Combining all rostered players’ projected BPR with predicted minutes gets an estimate of the team’s overall roster strength based on personnel. You can see each team's “Roster Rank” column on the Team Ratings page. The top 10 teams in roster strength are also shown below:

• Team performance in previous seasons - Though this is much less of a factor in a team’s preseason projection, it still does play an important role. Teams that often outperform their projected roster strength in past seasons are likelier to do so in future seasons. Teams with coaching changes are also handled differently from teams with returning coaches. Based on my analysis, when a team changes coaches, their next season's performance correlates very little with how they fared in the previous season. Roster strength is crucially important in these cases.

• Roster continuity - Teams with a higher percentage of returning minutes from the previous season are often slightly better than teams with a lot of roster turnover. This doesn’t play a huge role, but it does move the team projection slightly. I also added an estimate of each team’s percentage of minutes that will be played by returning players since my previous study on roster construction shows that teams who build their rotation more around returning players are more successful than teams with lots of newcomers, even with comparable roster strength.

Top 30 Teams

Let’s look at the top 30 preseason teams without further ado.

It’s not surprising to me in the slightest that Houston and Purdue are the top two teams in my preseason projections. Houston has an insanely good track record over the last couple of years, finishing in the top 4 at EvanMiya.com each of the previous five seasons. Their returning group this year is once again very solid, led by the trio of Milos Uzan, Emmanuel Sharp, and Joseph Tugler. Meanwhile, Purdue has assembled the top-ranked roster in the country, led by the preseason #1 player at EvanMiya.com, Braden Smith, who should be dominant this year.

Duke sits at third, which might surprise some people. Their roster clearly isn’t as talented as last season, but this year’s crop deserves more praise than they are getting. Freshman Cameron Boozer should be in the running for freshman of the year, and my model predicts big sophomore campaigns from both Isaiah Evans and Patrick Ngongba. These two were used in a limited fashion as freshmen but showed enormous promise and will be stepping into much bigger roles. Both land in the top 50 players in the preseason at EvanMiya.com.

Here are some other teams that my model seems particularly optimistic about:

• #5 Michigan boasts the 2nd-best roster in the country behind Purdue. Yaxel Lendeborg is the 2nd highest rated player in the preseason after Braden Smith.

• #8 Illinois has two star returners in Tomislav Ivisic (#13 at EvanMiya.com) and Kylan Boswell (#43), along with a hyped group of incoming talent from Europe. The more I dig into Big T’s numbers from last year, the more I am convinced he will be really special.

• #18 Ohio State might be the team in my top 25 that is the most surprising, but the Buckeyes had a great approach to their roster build this offseason. They return three massive contributors from last year in Bruce Thornton (#6), John Mobley Jr. (#78), and Devin Royal (#94). Getting your biggest players to return is really tough to do in the modern portal era. Thornton may not be the flashiest point guard in the country, but his efficiency as a scorer and playmaker is pretty impressive.

• #25 Vanderbilt once again has a chance to overachieve relative to expectations after being very smart about the players they have recruited this offseason. I was high on Vanderbilt last year (35th in the preseason) and was proven right to be optimistic. We will see if Mark Byington can live up to the analytical hype for a second year in a row.

• #27 Mississippi State is ranked this high in large part because of star point guard Josh Hubbard (#18 in the player ratings). Hubbard is the only guard in the country with a projected points per 100 possessions of over 34 while maintaining a projected turnover percentage under 2.5%. He rarely turns the ball over despite being asked to be the primary scorer and playmaker on the team.

• #28 Cincinnati being this high scares me a little bit, but on paper, the Bearcats have another solid roster. While they don’t have any bona fide stars, they really focused on getting some highly rated defensive players, which should pay dividends on that end of the court.

• #36 Miami has a top 10 incoming transfer class at EvanMiya.com and should be much improved under new head coach Jai Lucas.

On the flip side, here are some teams that my model seems to be down on relative to consensus:

• #13 St. John’s will be led by Zuby Ejiofor (#15 in the player ratings), but my expectations for some of the other key players are lower than most. Bryce Hopkins should be great when healthy, but it’s been a while since he was at his best. My model has some reservations about transfers Oziyah Sellers and Dylan Darling as well.

• #19 Louisville has a very impressive roster, ranked 6th best in the country. However, the expectations are that Pat Kelsey will be just as tremendous in year 2 as he was with an injury-laden roster in his first year at the helm. I certainly think this is a strong possibility, but my model is a little more skeptical. Some of this pessimism can probably be attributed to the Cards finishing outside the top 20 last year, but this could easily be excused due to all the injuries the team dealt with.

• #23 Texas Tech will be led by likely NPOY contender JT Toppin (#3), along with sophomore point guard Christian Anderson (#56), but I have some serious concerns about the roster depth for the Red Raiders. They only have five players ranked inside the top 500 players in the preseason.

• Some other teams my model is lower on include #35 Wisconsin, #43 San Diego State, #46 Virginia, and #48 Washington.

Model Updates

I made some significant improvements to the Bayesian Performance Rating model that guides all player evaluations at EvanMiya.com, which I will dedicate a whole article to in the coming weeks. You will find updated player ratings for every season going back to 2009-10 on the site.

What’s Next?

As we get closer to November, I’ll be posting more preseason content here on my blog as well as on Twitter, Instagram, and TikTok. I also have a new small merch store that you can find a link to on the main site.

The new tool takes an input of any list of players and predicts roughly how good a team with that roster would be in the preseason. The roster grade is presented as a range of plausible national rankings (such as 26th to 32nd) for the overall team ranking, offensive ranking, and defensive ranking. These rankings are based on where a team with the chosen roster would be ranked in a typical preseason.

Even if a roster isn’t complete, it can still be evaluated based on the players the team has so far. You can also include other potential additions to see how certain players would improve the team, or you can pick generic placeholder players, such as “Big 12 Starter”. When next offseason rolls around, the Roster Outlook tool will be the easiest way to assess rosters on the fly and plan for upcoming moves.

I would also strongly recommend these databases from Jim Root (Three Man Weave) and Jeff Borzello (ESPN) when it comes to tracking offseason roster movement.

It’s important to note that the tool only considers the talent on the roster, not the coach or the team's recent track record. For example, Houston’s roster talent would be analyzed based solely on the individual players, without accounting for Kelvin Sampson’s impact on the team.

To give a sneak preview of the Roster Outlook tool, I’ve picked one team from each major conference, along with one interesting team outside those conferences.

Big Ten: Purdue

Purdue had an incredible offseason, retaining almost all of their key players from last season, including Braden Smith, Trey Kaufman-Renn, and Fletcher Loyer. Based on current projections, Smith is likely to be the highest-rated player in the country in the preseason at EvanMiya.com. With so many other strong supporting players, including incoming big man Oscar Cluff and sophomores CJ Cox and Gicarri Harris, there’s a decent chance Purdue is the #1 ranked preseason team on our site. Purdue’s roster continuity is a significant reason why: returning players are expected to play 77% of the team’s minutes next season. The Roster Outlook tool says that Purdue’s roster would typically have a preseason ranking between 1st and 4th.

Big 12: Kansas

Kansas won’t be in contention for the top spot in the preseason AP poll this year, but its roster still has a lot of potential. Darryn Peterson is expected to be the star on this year’s team, but I anticipate Flory Bidunga having just as big of an impact, especially on the defensive end. The Jayhawks’ defense has a higher expected rating than the offense, with the overall team roster being ranked between 15th and 25th in a typical preseason.

Big East: UConn

Alex Karaban’s decision to return to Storrs for his senior year took this UConn roster from good to great. UConn’s roster ranking range is between 4th and 9th, which should be a massive improvement over last year’s underwhelming squad. If Karaban had chosen to leave for the NBA, Connecticut’s ranking would have been about 10 spots worse. Silas Demary and Malachi Smith are also really significant additions through the portal.

SEC: Tennessee

Tennessee’s roster features an interesting mix of veterans and younger players with considerable potential. The biggest win for the Vols this offseason was acquiring transfer Ja’Kobi Gillespie from Maryland, the third-best player in the portal, according to EvanMiya.com. Along with Felix Okpara, Gillespie will provide some much-needed experience to pair with Tennessee’s exciting young talent, headlined by freshman phenom Nate Ament. The Roster Outlook tool says that this roster would typically have a preseason ranking between 18th and 26th.

ACC: NC State

I’d argue that no team has had a flashier offseason than NC State. New head coach Will Wade quickly hit the recruiting trails hard, putting together the third-best incoming transfer class in the entire country, according to EvanMiya.com. Darrion Williams is the big catch, as Williams is the best player to enter the transfer portal this offseason, according to EvanMiya.com. Williams is good enough to be a first- or second-team All-American this year. The rest of NC State’s roster is probably underrated by most as well. Tre Holloman, Ven-Allen Lubin, Terrance Arceneaux, and Quadir Copeland are all substantial additions. NC State’s roster ranking range is pretty wide: between 5th and 20th best in the country.

It’s interesting to see how much of an impact the addition of Darrion Williams has on NC State’s outlook for 2025-26. If you submit NC State’s roster without D5, the team’s roster ranking range drops from 5-20 to 33-39.

Other Conferences: Gonzaga

Gonzaga lost some key players to graduation this offseason: Ryan Nembhard, Nolan Hickman, Ben Gregg, and Khalif Battle. However, Graham Ike is back and will be a prolific scorer once again. Braden Huff will also have a chance to be fully unleashed this year and will be a force to be reckoned with on the offensive end. Between Ike, Huff, Innocenti, Smith, and Diagne, the Zags’ returning players are expected to play 64% of the team’s minutes this year, a very healthy number. Mark Few also has Tyon Grant-Foster and Adam Miller transferring in, who are likely to play significant roles. The Roster Outlook tool says that this roster would typically have a preseason ranking between 10th and 22nd.

Accessing The Tool

The new Roster Outlook tool is available to Premium subscribers at EvanMiya.com. In addition to this feature, subscribers get access to a ton of other fabulous things, including:

• Player skill projections and grades

• Lineup grades, strengths, weaknesses, and optimization tools

• Matchup previews and predicted score explanations

• Keys to victory for every team

• Hypothetical transfer portal lineup projections

• Game injury reports and injury line adjustments

There are over 100 Division 1 programs with subscriptions to EvanMiya.com. In addition, thousands of regular college basketball fans are also signed up because the public tiers are so affordable (just $29.99 per month for the Premium tier or half that when paid annually). Come join the fun and sign up over at EvanMiya.com!

Power-conference programs often have the resources to build a CBB roster in any manner they choose, but the same can’t be said for the rest of Division 1. It’s harder than ever for mid-major programs to have control over their rosters. The best players from these schools often get poached through the transfer portal, but even rotation players think the grass could be greener elsewhere. How can these programs still find regular success?

Today’s study focuses on how to build a tournament-caliber team for programs outside the high-major level. This includes all teams outside the ACC, Big 10, Big 12, Big East, and SEC, along with Gonzaga and Memphis. I decided to include programs from conferences that can often earn at-large tournament bids, such as the Atlantic-10, Mountain West, American, and WCC. However, some distinctions will be made for these conferences when appropriate.

In this article, we will answer many of these roster-building questions that are pertinent to mid-majors:

How important is roster continuity?

How much does experience matter?

Where can you find stars at the mid-major level?

How can you convince key players to stay for multiple seasons?

Where can you find the best transfers?

Mid-Major Roster Construction Best Practices

We will start with the main conclusions of this 4-year study and then explain each one more in-depth, along with the accompanying research:

1. Retain key talent as much as possible

2. Look for down-transfers from higher levels

3. Limit reliance on freshmen

Studying Recent Trends

Let’s dig deeper into the roster construction of mid-major teams over the last four years. I chose to limit the study to the 2021-22, 2022-23, 2023-24, and 2024-25 seasons because they most closely represent the current landscape of college basketball, with plenty of immediately eligible transfers and NIL playing a massive factor in recruitment. In some cases, I limited this to just the last two seasons when appropriate.

The main variables that I took into consideration for this study were:

1. Preseason roster score ranking1

2. The percentage of minutes played during the season by returning players

3. The percentage of minutes played by new transfers

4. The percentage of minutes played by freshmen

5. The average years of Division 1 experience on the roster coming into the year

There are different ways of defining what “success” looks like for mid-majors, but I decided to go with what most people care about: making the NCAA tournament. When you compare the roster construction of mid-majors that made the tournament to those that missed the tournament over the last several years, there are some notable differences:

Teams that played in March Madness were much more reliant on returning players, especially at the expense of freshmen. Over 64% of the minutes on tourney-teams went to players on the roster the previous season, almost 10% higher than teams that missed the dance (55%). Meanwhile, just 11% of the minutes for tourney-teams were allocated to incoming freshmen, a much lower mark than for all other teams.

The level of experience on the roster was also higher for teams that played in March Madness, at 1.99 average years of D1 experience, compared to 1.66 for all other teams. Notably, there wasn’t a significant difference in the importance of transfers to these two groups of teams.

If you just isolate this data to mid-majors from 2024-25, teams overall weren’t able to return as many players as in previous years, but those that made the tournament still relied significantly more on returning players than those that missed the dance:

To look at specific success cases, the table below shows the roster composition data for every mid-major team that has won at least one game in the tournament since 2022. Of the 25 teams, 18 gave at least 50% of their minutes to returning players. Perhaps more importantly, 20 out of the 25 teams played freshmen for under 10% of their minutes:

Mid-majors that rely more on incoming transfers can find tournament success. All of these teams that won a game in the tournament and had transfers play over 40% of their minutes: Drake and McNeese in 2025, Utah State, Dayton, Washington State, James Madison, and Grand Canyon in 2024, and Fairleigh Dickinson, Murray State, and New Mexico State in 2022.

However, most of these teams were loaded with experience on the roster. When taking a weighted average of the years of experience on the roster based on playing time, 84% of the teams that have won a game in the tournament averaged over 1.8 years of Division 1 experience coming into the year.

Where Do Mid-Major Stars Come From?

Winning in March often requires having a star player or two, but finding and retaining star talent at the mid-major level is hard. To find out where the best mid-major players get recruited from, I looked at the top 3 players on every mid-major team according to Bayesian Performance Rating at EvanMiya.com and split them into four recruitment categories:

• Incoming transfer

• Incoming freshman

• Returning player, originally recruited by that team out of high school

• Returning player, originally recruited through the transfer portal

The table below shows where the “star” mid-major players are recruited from, comparing teams that made the tournament to teams that missed it:

The big takeaway is that 50% of the star players who made March Madness were recruited by that mid-major school out of high school who stayed for multiple years. So many of the best mid-major players were developed the old-fashioned way: staying with the same team and improving year after year.

Even this last year, many of the best mid-major players in college basketball were guys who originally signed out of high school. The table below shows the top 25 mid-major players in 2024-25 according to Bayesian Performance Rating and their recruitment type. Out of the top 25 players, 11 of them were originally high school recruits:

For every player who chooses to return to his school and be a leading star, more players jump ship through the portal. Every mid-major coach wants to retain the team’s core players each season, but they often struggle to do so since they usually can’t pay as much in NIL as a bigger program might be willing to do.

In my opinion, the pitch to keep core players around another year is to convince them not to transfer to a higher level till they are a bona fide star and can truly capitalize on their market value.

If you look at the best mid-major player on every team this year, a lot of them were returners from the year before, but were not the best player on the team in the previous year. In fact, if you take the top 25 mid-major players who led their team in BPR in 2025 as returning players2, there was a significant improvement from the 2024 season to this season. On average, these players were around the 2nd or 3rd best performer on their roster in 2024, ranked around the 396th best player in the country that year. They all became the best player on the team in 2025, improving to the 124th best player in the country on average.

It might be difficult in a lot of cases for a mid-major to stop their best player from transferring. But if a program can retain the guys next in line as the 2nd, 3rd, or 4th best player on the team, they often improve tremendously and can be a catalyst for huge team success before taking advantage of their higher profile and transferring.

Finding The Best Transfers

While retaining as much talent as possible should be the first priority for mid-majors, identifying great fits through the transfer portal is also crucial. Coaches constantly seek the right avenues to find players through the portal, hoping to find a solid contributor or a hidden gem with untapped potential.

The table below shows the top 25 mid-major players from the last two years, according to BPR, who came to that school originally as a transfer (D1 only). I have included each player’s previous team, along with that team’s conference level, split into “High Major”, “Mid Major+” (A-10, American, Mountain West, WCC, Missouri Valley), and “Mid Major”:

A majority of the best D1 transfers to the mid-major level have come from high-major schools (13 out of the 25), and just 2 of the transfers came from a true mid-major or non-D1 school. This shows that the most reliable transfers are players who are down-transferring or staying at the same level of competition. While there are some successful examples of players from low-majors (and outside Division 1) moving up to a higher level, it’s rare for these players to make as big of an impact as transfers from better levels.3

Additionally, many of the best high-major to mid-major transfers fit a particular mold: They were 3 or 4-star recruits out of high school who didn’t find consistent playing time. Of the 13 successful high-major transfers in the previous table, only 2 were 3-stars or better coming out of high school.

Many of the best mid-major incoming transfers have been high-major rejects who couldn’t crack 20 minutes a game at their first school. I pulled a list of some of the most successful players at the mid-major level who were originally at high-major schools but weren’t able to find regular playing time:

There are some big names on the list, including Vlad Goldin (originally at Texas Tech before going to FAU), Jaedon LeDee (TCU to SDSU), Jaelen House (Arizona State to New Mexico), and Drew Pember, who averaged under 3 minutes a game at Tennessee before becoming the two-time Big South player of the year.

The Freshman Payoff

As mentioned previously, teams that make the tournament as mid-majors aren’t usually that reliant on freshmen talent. In the last 4 seasons, if you take the top 5 players from every mid-major team that won a tournament game, just 6% of those players were freshmen.

However, recruiting quality freshmen out of high school is still very important, especially if those players can be retained for multiple years. Many of the best mid-major stars came in as freshmen and improved over multiple years at the same school. While you probably can’t expect them to be major contributors in their first year, the hope is that they will be among your best players in future years.

More Tools

When evaluating players each offseason and deciding what players to keep and what players to pursue in the portal, there are many tools at EvanMiya.com dedicated to comprehensive player assessments and projections. Over 100 Division 1 programs have subscriptions and actively use the site for player evaluations, lineup analysis, game-scouting, and more.

Along with transfer portal ratings for all D1 players based on their predicted value next year, every player has individual player skill projections and grades for all key areas.

You can learn more about these player projections here, along with the lineup optimizer tool here. I also recently released the Front Office Suite tool, designed to help teams and GMs build rosters and get an NIL value for every player, adjusting to the current market.

It’s an even bigger deal at the high-major level, as teams with huge NIL budgets can overhaul any roster with new players through the transfer portal if they are willing to invest enough money. There were six high-major teams in 2024-25 that returned zero minutes from their previous roster, a feat never attempted in years past.

Teams are utilizing the transfer portal more than ever this offseason. Last year, there were 1653 completed transfers in Division 1, according to data at EvanMiya.com. Less than a month into the portal being open this offseason, there are already more than 2,300 student-athletes in the portal. The changing climate has sparked numerous discussions on the best approach to building rosters in the modern portal era. While the portal provides every coach with the opportunity to completely rebuild in a single summer, questions have been raised about whether transfer-reliant teams are actually effective.

Three of this year’s Final Four teams, all 1-seeds, were comprised mostly of players returning to their team from the previous year. Houston, Florida, and Auburn all had at least 69% of their minutes being allocated to guys coming back for another year.

Last offseason, I published an article titled “How To Build A Roster In The Modern Portal Era,” which examined recent roster construction trends and outlined a strategy for success for teams seeking to build championship-level squads. Today, we will revisit that study, updated with new data from the 2025 season.

In this article, we will answer many of the roster-building questions that are being talked about in today’s climate:

Can you have too many transfers?

Is transfer recruitment more critical now than high school recruitment?

How valuable are returning players?

Do you need freshman talent to be elite?

Should you prioritize returning players over transfers?

Is there a “formula” for roster composition?

For the purposes of today’s discussion, we will only talk about high-major teams, as the resources at mid-major schools dictate that they approach roster construction in a very different manner from high majors. I will dedicate a future article in the coming weeks to applying these concepts for mid-majors.

Roster Construction Best Practices

We will start with the main conclusions of this 4-year study and then explain each one more in-depth, along with the accompanying research:

1. Fill your roster with good basketball players

2. Returning players should account for at least 50% of the playing time

3. Prioritize recruiting players who will play at least 2 seasons for you

Talent Matters

The most important (and obvious) strategy in roster construction is always to fill your roster with skilled and talented players. It doesn’t take a statistical study to convince anyone of this point, so I won’t spend much time on the topic.

The strength of a team’s squad in the preseason will always be the strongest indicator of success during the season. As part of this study, I calculated a preseason roster score for every team by taking the preseason BPR projection for each player on the roster and summing them up, weighting each player based on the number of minutes they are likely to play. These minute predictions are based on how each player ranks relative to other teammates in their BPR projection. The best players on a team typically are predicted to play around 30 MPG, and the worst players in the rotation are expected to play around 10 MPG.

The graph below shows how every team’s preseason roster scores against their end-of-season team efficiency rating at EvanMiya.com over the past four seasons.

Unsurprisingly, there’s a high correlation between roster score and team performance during the season. If a coaching staff is deciding between two players, one of whom is definitely a better player than the other, it’s always the best choice to go with the more talented player, no matter what avenue of recruiting that player comes from (returning player, transfer portal, or high school).

A Four-Year Study

With the obvious out of the way, let’s dig deeper into the roster construction of high-major teams over the last four seasons. I chose to limit the study to the 2021-22, 2022-23, 2023-24, and 2024-25 seasons because they most closely represent the current landscape of college basketball, with numerous immediately eligible transfers and NIL playing a significant role in recruitment.

The main variables that I took into consideration for this study were:

1. Preseason roster score ranking

2. The percentage of minutes played during the season by returning players

3. The percentage of minutes played by new transfers

4. The percentage of minutes played by freshmen

The following table shows how the previous three champions fared in each category. Notice that each title winner used returning players for over 50% of the available minutes.

Since it’s impossible to make sweeping conclusions on roster construction from just four teams, let’s expand our pool of “successful teams” to all of the 1 seeds from the past four seasons. Every team had a preseason roster rank within the top 30, with 13 of 16 having top-15 rosters. Almost all of the 1 seeds leaned mostly on returning players, as only three teams had returning players play less than 50% of their team’s minutes (Alabama 22-23, UNC 23-24, Duke 24-25).

This past season, all four 1-seeds made the Final Four, with Houston (82% of minutes played by returners), Florida (70%), and Auburn (69%) all building their teams around a key group of players coming back for another year. Duke’s team was constructed very differently, but had arguably the most talented roster in the country featuring three likely lottery picks in the 2025 NBA draft.

Let’s expand our list of successful teams a bit further and compare all teams that finished in the top 10 at the end of the season (all of which made the tournament) to all high majors that missed the NCAA tournament entirely.

The differences are pretty stark. Top 10 teams, on average, had returners play about 60% of the available minutes during the season, while teams that missed the tournament were only at 48%. The reliance on transfers was significantly lower for top 10 teams, at just 24% of the minutes, compared to 38% for high majors that failed to secure a tournament berth. Top 10 teams also took 1.3 fewer transfers in the offseason on average than teams that missed the tournament. Looking at just this past season (2024-25), high-majors who missed the tourney only played returners for 39% of their minutes, compared to 57% for top 10 teams.

Hitting At Least 50% Minutes From Returning Players

If you compare two rosters that are of similar talent in the preseason, the team that is structured around returning players typically performs better than the team that is structured around mostly new transfers and freshmen.

To illustrate this, the graph below categorizes every D1 team from the last four seasons into groups based on their preseason roster score ranking. For example, all teams ranked between 110th and 70th are grouped together, while those ranked in the top 35 are in another group. When you look at the average end-of-season ranking at EvanMiya.com for teams within each group, the teams that gave returning players at least 50% of the team’s minutes during the season always finish better on average than the teams that gave a majority of their minutes to new players.

For example, among teams that had top 35 preseason roster scores, those that played returning players at least 50% of the time typically finished around 18th in the country at the end of the year. However, teams that played returners less than 50% of the time finished around 26th on average, 8 spots worse. This difference is evident in almost every grouping of teams, regardless of their preseason roster talent.

Interestingly, the percentage of minutes played by freshmen doesn’t seem to impact whether a team performs well or not. In my opinion, teams built around transfers struggle more often because transfers often fill a void for a potential returning player who was not retained. Teams with many transfers will likely have very few returning players, and vice versa.

The same is not the case for rosters studded with freshmen. These teams are more likely to still have a good number of returning players, unlike teams composed entirely of transfers.

Broad Conclusions

Based on the results of this study, the most crucial factor for team success, beyond prioritizing good players, is maximizing the role that returning players will play. Having returners account for at least 50% of the minutes played is a safe benchmark, but teams with a higher percentage are more likely to perform well. As Trilly Donovan often says, “Sometimes the best gets are the ones you already have.”

A secondary takeaway is this: Teams that prioritize recruiting (or retaining) players who will play for them for multiple years will reap the benefits down the road. The actual value of a freshman or transfer can often come in the 2nd or 3rd year that he plays for the same team. Final Four standouts like Johni Broome, Walter Clayton, and LJ Cryer this past season were in their 2nd or 3rd year with their team after transferring. Given how much more success teams with returning players have, it is wiser to recruit a player who is more likely to play multiple years at your school than a similarly skilled player who will just play one season and graduate or transfer away.

Case Studies

One of the most interesting parts of this research is analyzing teams that take some of these roster-building strategies to the extreme. Here, we will examine outlier teams from several categories and provide insight into whether these approaches to roster construction work.

Running It Back: Teams Who Return Maximum Players

The table below shows the high majors from the last two seasons with the highest percentage of minutes played by returning players. All of these teams had returners make up at least 75% of the possible minutes on the court:

Twelve of the fifteen teams made the tournament, 10/15 finished in the Top 30 at EvanMiya.com, and two made the national title game (Purdue 23-24 and Houston 24-25). Teams that “run it back” typically have high, safe floors. They don’t usually underperform and are likely to have pretty good seasons, if not better.

Start From Scratch: Teams Who Return Minimum Players

The table below shows the high majors from the last four seasons with the lowest percentage of minutes played by returning players. Every team had returning players play less than 16% of their team’s minutes:

The results are pretty disappointing for this group. Only 6 teams out of the 20 made the tournament, and only three advanced past the first round. Almost half of the group finished outside the Top 100 at EvanMiya.com. Of the six high-majors this past season who had entirely new rosters, only Kentucky and Louisville made the tournament. It takes an excellent coaching staff to get the most out of a roster comprised entirely of new faces.

Young And Talented(?): Teams Who Maximize Freshmen Minutes

To analyze teams that rely heavily on freshmen, I will break them into two categories.

First, we will look at the teams with top 30 preseason roster scores with the highest percentage of minutes played by freshmen. These teams typically featured elite five-star freshmen and were likely to qualify for the tournament. Every school on this list had freshmen account for 30% of the minutes or more, with Duke (22-23), Kentucky (23-24), and Duke (24-25) exceeding 50%.

These teams, boasting highly touted freshmen classes, have fared quite well, with all ten teams making the tournament. What’s interesting, though, is that while the freshman talent was sky-high, these teams also had some key returning players to pair with freshmen to make it work. Among this group, 39% of the minutes played were by returning players and 44% by freshmen. Generally, it’s hard for teams with big-time high school prospects to fail, especially if there are equally skilled returning players to match with them.

Next, we will examine schools ranked outside the top 30 in preseason roster scores and look at the 10 teams with the highest burden on freshmen from that group. Every team in this category played freshmen for at least 37% of the team’s minutes:

The results aren’t pretty for teams that rely heavily on freshmen but aren’t among the nation’s elite rosters. Just 30% of this group made the tournament, and 5 out of the 10 teams finished the year outside the top 75 in the country. While the average percentage of freshmen minutes from this group was 46%, just 30% of the possible minutes were played by returning players.

The 2025 Rutgers team, featuring Dylan Harper and Ace Bailey, is an interesting data point here because the Scarlet Knights were widely regarded as a top 25 team in the preseason. While Harper and Bailey were ultra-talented, the other rotation pieces around the freshman duo left a lot to be desired, which is why they were ranked 70th in the preseason at EvanMiya.com. Ultimately, they performed at pretty much that exact level, finishing the year 76th in the country.

What About Mid-Majors?

Stay tuned for a future blog article in the coming weeks on optimizing roster construction for teams outside of the high-major ranks.

In this bracket primer, we will discuss:

• Who are the true title favorites?

• Which teams have the most value based on where they are seeded?

• Who are this year’s cinderellas?

• What can the “Kill Shot” tell us about tournament success?

Without further ado, let’s dive in!

Predicted Efficiency Landscape

The place to start is by examining every tournament team’s predicted team strength at EvanMiya.com, comparing offensive and defensive efficiency. These ratings are adjusted for current injuries, which means that teams like Duke (Cooper Flagg questionable), Alabama (Grant Nelson), Texas Tech (Chance McMillian, Darrion Williams), Iowa State (Keshon Gilbert out), and Kentucky (Lamont Butler) are properly docked for the uncertainty of their players’ availability. You can find complete injury reports, updated daily, in the Matchup Previews tool at EvanMiya.com.

The graph below shows every team’s predicted offensive rating and defensive rating, with teams split into tiers based on their tournament potential:

Title Favorites

As mentioned in a previous article, teams heading into the tournament in the “Title Favorites” tier have won the last 9 national titles. This year, five teams qualify: Duke (1 seed), Houston (1), Florida (1), Auburn (1), and Alabama (2).

The four 1-seeds, in particular, are very strong this year. According to bracket simulations at EvanMiya.com (at the bottom of the article), there is a 68% chance that a 1-seed will win the title, which aligns with the actual track record of 1-seeds. Since 2000, 17 of the 24 national champs (71%) have been 1-seeds. Alabama is also a strong team, but they were handed a tough draw being in the same region as Duke, with just a 4.8% chance of winning the title.

Tournament Team Strength vs Bracket Seed

To better understand at how these team ratings compare to each team’s seed, I present the graph below, which highlights teams that are much stronger or weaker than usual for their seed. This is really where we can find value in the predictive analytics. Teams highlighted in orange are considered “underseeded” and advance 59% further than other teams with the same seed.

This year, 12 teams qualify as underseeded due to being much better than usual for their seed. Duke also qualifies, but they can’t get a better seed than they already have. Of these teams, the ones that really stick out are 8-seed Gonzaga, 11-seed VCU, and 11-seed North Carolina, who all have a predicted team strength much better than their seed.

This year, the group of 11, 12, and 13 seeds is particularly strong. VCU has a 50.2% chance of beating 6-seed BYU in their first-round matchup, Colorado State has a 53% chance of beating 5-seed Memphis, UC San Diego has a 45% chance of advancing past 5-seed Michigan, UNC has a 54% chance of beating 6-seed Ole Miss (if they can get past San Diego State first), and even 13-seed Yale has a 30% chance of advancing past 4-seed Texas A&M.

I’m very torn on the potential 8-seed Gonzaga vs. 1-seed Houston matchup. Gonzaga is a top-10 team, according to EvanMiya.com, but Houston is the 2nd best team in the field. The Matchup Preview tool, which lets you preview any possible matchup between teams, shows Gonzaga would only have a 33% chance of beating Houston in that game.

Maryland has the best chances of advancing to the Final Four (11.6%) of any team not in the top three seed-lines. If they advanced past Grand Canyon, they would face the winner of Memphis/Colorado State, which would make them comfortable favorites either way.

On the opposite side of the coin, some overseeded teams are likelier to come up short in the tournament, compared to usual expectations for their seed: 3-seed Kentucky (dealing with injuries), 5-seeds Michigan, Oregon, and Memphis, and 6-seed Ole Miss.

The Kill Shot

This wouldn’t be an EvanMiya.com article without analyzing the “Kill Shot”, any time a team scores 10 or more unanswered points in a game. Teams that go on many of these 10-0 scoring runs are more able to erase a deficit quickly and can also easily close the door on an opponent in a tight contest. Similarly, teams that limit opponents from having “Kill Shots” are more likely to hold on to a lead instead of allowing another team back into the game.

Before we jump into analyzing individual tournament teams, here are the stats about scoring runs that you need to know:

• A team with at least one 10-0 run wins 71% of the time

• A team with more Kill Shots than the opponent wins 81% of the time

• A team with 2 Kill Shots in a game wins 88% of the time

• A 15-0 run leads to a win 86% of the time

• A 20-0 run leads to a win 91% of the time

The graph below shows the landscape of tourney teams in 2025 in terms of scoring runs per game and scoring runs conceded per game, adjusted for opponent quality:

As shown on the graph, I like to put teams in four categories: “Strong” teams, “Streaky” teams, “Least Streaky” teams, and “Suspect” teams.

All the major title favorites fall in the “Strong Teams” category, but Texas A&M, Iowa State, Michigan, Marquette, Memphis, Missouri, UCLA, and Michigan State are here too.

Duke is the best team overall at going on 10-0 runs and not giving them up in return. Some of their biggest games this year have featured a big scoring run at the right moment.

The big standout is Yale, which is rated as the best team in the country at going on 10-0 runs, even after adjusting for opponent strength. This could spell trouble for Texas A&M in their first-round matchup, although A&M has only given up 9 Kill Shots all year.

Alabama didn’t quite make the cut for the top tier, but their Adjusted Kill Shots Conceded per game is lower than any other team in the country. It’s hard to string a bunch of scoring possessions together against them without the Crimson Tide scoring in return.

This is where we can really have some fun with this metric. The six tournament teams in the “streaky” category are all very capable of going on scoring runs… but they also give them up just as quickly. Games with these teams are often wild, with quick lead changes back and forth. No game with a “streaky” team is over, win or lose, until the very end.

There are two main takeaways for teams in the “streaky” category. First, they are vulnerable as a favored team but have more upset potential in games where they are the underdog. Because scoring spurts can happen at any time in either direction, their games are very unpredictable. A team like Illinois could either flame out in the first round or get hot and go deep in the tournament. And you really don’t know which one will happen.

The other takeaway is that live spreads for games with these teams could be less accurate because any lead could be vaporized quickly. If a team like Iona is down 10 at halftime, they will be more likely to mount a comeback than most other teams of similar quality.

In contrast to the last category, we now look at teams in the “least streaky” group, which are more boring by nature. These teams have not had as many 10-0 runs this season, but they rarely give them up. Liberty has only given up 3 Kill Shots all season!

Teams in the “least streaky” category are pretty trustworthy when they have a lead since they are not very likely to allow the opposition back into the game quickly. However, teams like Ole Miss and Utah State often struggle to overcome a deficit quickly. Look to rely more on these teams when they are favored, but be cautious to count on them winning several games as an underdog.

Lastly, we touch on the “suspect” group, which comprises teams prone to giving up scoring runs to the opposition while struggling to counter on their own. There are a lot of bad teams on this list, but it also notably includes Vanderbilt, who has given up a ton of 10-0 runs against high-major opponents.

Now, what does all of this mean for your team’s chances of reaching a Final Four?

Let me first state that while the Kill Shot is a helpful tool, it can’t be used in a vacuum to determine the best teams in the country. Instead, I view it as a useful companion to other metrics to provide more nuance about the characteristics of each team.

With that said, since the 2018-19 season (which is as far back as my Kill Shot database goes), heading into the tournament, every national champion has had an Adjusted Kill Shots Per Game of at least 0.47, an Adjusted Kill Shots Conceded Per Game under 0.34, and an Adjusted Kill Shots Margin of at least 0.28.

There are 11 teams that qualify using this year’s statistics: Duke, Houston, Florida, Auburn, Alabama, St. John’s, Michigan State, Missouri, Texas A&M, Marquette, and Memphis.

In general, I have found that preventing 10-0 runs is the more predictive stat of teams that advance the furthest in the tournament. While it’s great to be able to put up scoring runs in a hurry, the best teams are the ones who can hold a lead and not quickly allow an opponent back in the game.

Matchup Previews

To dig more into the individual matchups, the Matchup Preview Tool at EvanMiya.com lets you analyze any hypothetical matchup in the tournament. It gives you score predictions, injury adjustments, other matchup-specific adjustments, a team comparison, a roster breakdown, and a daily updated injury report.

This preview shows the potential top-10 matchup in the 2nd round between Houston and Gonzaga:

Tournament Probabilities By Region

The tables below show each team’s chances of advancing to each round of the tournament, which you can also find at EvanMiya.com, updated daily as the tourney progresses.

In 2024, I introduced a new method for rating teams in college basketball called Relative Ratings. While many readers are probably familiar with how adjusted team efficiency metrics work, Relative Rating has a few distinctions:

1. Teams are ranked based on who is predicted to win each possible head-to-head matchup. If a team is ranked number one, they are predicted to beat every other team in the country in a theoretical neutral-court matchup. These matchup predictions are calculated using team-specific contextual variables to determine likely head-to-head game winners.

2. Several matchup-specific variables are used to obtain predictions for each potential game, including opponent strength, game pace, recent form, and consistency measurements.

You can read more about how it all works here.

As it turns out, the pre-tournament Relative Ratings at EvanMiya.com are very predictive of team success in March Madness. In particular, there are two ways in which these team ratings will give you a massive edge in your bracket pools: picking potential title-winners and identifying under-seeded teams.

Picking Title-Winners

We can easily identify which teams have a good shot at winning the national championship by using pre-tournament Relative Ratings from the previous seasons (which you can find on the Team Ratings page at EvanMiya.com). The last nine champions have had a Relative Rating within 6.5 points of the highest-rated team heading into the tournament. The table below shows each of these title-winners:

For example, in the 2024-25 season, according to Relative Rating, the highest-rated team heading into the tournament was Duke, with an RR of 35.9. There were three other teams with 6.5 points of Duke: Houston (34.0), Florida (32.9), and Auburn (30.1). Florida ended up winning that tournament.

When the bracket comes out on Selection Sunday this March, I will identify the teams that meet that criteria. To give a little wiggle room, I will select teams within 7 points of the best team in the sport. In a given year, between 5 and 10 teams typically qualify, but there have been as few as 3 and as many as 12. Today, the list of qualifying title favorites is moderate: Michigan, Duke, Arizona, Florida, Houston, Illinois, Purdue, and UConn. Iowa State is 7.5 behind Michigan, just missing the threshold. This will likely change some by Selection Sunday, but that group of teams have distanced themselves so much from everyone else that they have a much stronger chance at a title than the rest of the country.

Identifying Under-Seeded Teams

The teams that make the most noise in March are often the ones who are much stronger analytically than the other teams on their seed line. We can use Relative Ratings to identify these teams each season heading into the tournament.

Using some modeling, we can find a typical team rating for every seed and then identify teams significantly stronger than expected at their seed line. As an example, here is every team’s pre-tournament seed in 2024, compared to their Relative Rating:

The solid black line represents a typical Relative Rating for that seed, and the dotted black line shows the threshold for under-seeded teams. In particular, I have found that teams with a pre-tourney Relative Rating at least 2.1 points higher than expected for their seed perform much better in the tournament.

Here are the stats you need to know about these under-seeded teams:

• Under-seeded teams advance 69% further than other teams on the same seed line, on average.

• About 21% of the teams in the tournament qualify as under-seeded.

The graph below shows the average number of tournament wins by seed, comparing the under-seeded teams (in orange) to all other teams (in blue):

Calculating the typical Relative Rating per seed is specific to each year, so I will put out a graphic on “Bracket Monday” highlighting those teams for 2026.

Pre-Tourney Graphs By Year

To highlight the success of this method for selecting teams that will advance further in the tournament, I am including a graph of every recent tournament and a table showing how each under-seeded team performed in that bracket.

Admittedly, the 2025 version of this analysis did not do particularly well. Teams labeled as “under-seeded” won only 10 games, which was less than the 12.4 typically expected for teams on those seeds. However, in the past 10 tournaments, the group of teams identified as under-seeded have collectively out-performed seed expectations in eight of the ten tournaments.

2020 (no tournament)

One commonly used resume sorting metric is problematic: the quadrant system. The premise of the quad system is simple: a team’s wins and losses are divided into four quadrants based on how difficult that game is deemed to be, using a combination of the opponent's NET ranking and whether the game took place at home, on the road, or at a neutral site.

On the surface, a team’s quadrant records provide a good summary of the types of wins and losses that a team has. For example, here are Louisville’s quadrant records as of Wednesday, January 29th, courtesy of WarrenNolan.com:

An experienced resume evaluator can immediately see that Louisville has performed pretty well in its toughest games (.500 in Q1) while handling all other easier games without a hitch (11-0 in the non-Q1 games). In fact, just five teams have more Quad 1 wins than Louisville: Auburn, Oregon, Iowa State, Kentucky, and Texas A&M.

The most glaring issue with the quadrant record system is the sensitivity of the quadrant cutoffs. A Quad 1 loss can turn to a Q2 loss overnight, just due to that opponent moving from the 75th to 76th in the NET without even playing a game. The cutoff points are arbitrary, picked to be nice numbers divisible by 5 so that they are easy to remember:

Furthermore, all wins within the same Quadrant are treated equally, which is incredibly unfair. A win on a neutral against Bradley counts as a Q1 win, just like a win on the road at Auburn would. These hard-set quadrants make sorting wins and losses into categories easy, but you lose so much valuable information in the process.

In summary, the quadrant system aims to explain how good a team’s wins and how bad a team’s losses are, but the execution is flawed. I’ve been told that the selection committee is much more nuanced in the way it evaluates these wins and losses, but for someone who is taking quadrant records at face value, it can be misleading.

Resume Quality at EvanMiya.com

To improve this, I have created a new resume metric at EvanMiya.com called Resume Quality, which divides your body of work into two numbers: Win Quality and Loss Quality. Similar in nature to Wins Above Bubble (BartTorvik.com), Resume Quality determines how difficult each game on a team’s schedule is based on how a team on the at-large cutline would fare in those games. The real value of this metric comes in splitting up wins and losses into several categories:

Win Quality: This measures how good your wins are based on the difficulty of those games. A team with a high win quality has won many tough games.

Loss Quality: This measures how bad your losses are based on the difficulty of those games. Having a loss quality close to zero (loss quality is negative) means that you don’t have many losses, and/or your losses came against good opponents. A lower loss quality often indicates that you have lost some games you should have won.

Resume Quality: This is the sum of Win Quality and Loss Quality. It’s also equal to your number of wins minus the expected number of wins that a bubble team (defined as a team good enough to get the last at-large bid) would have against your schedule.

The table below shows the top teams in Resume Quality, which can be found on the Team Resumes page at EvanMiya.com:

You can also hover over any team’s Win Quality or Loss Quality to see the value of their best wins and worst losses. For example, here are Auburn’s best wins, with their Win Quality values:

How It’s Calculated

The difficulty of each game is determined based on how likely an average bubble team (defined here as a team right on the tournament at-large cutline) would be to win against that opponent at that venue.

For example, Mississippi State lost at Auburn this season. According to the current team ratings at EvanMiya.com, the probability that an average bubble team would win at Auburn is just 14%. If Mississippi State loses, its loss quality decreases by 0.14. However, if It wins, its win quality increases by 0.86, which is equal to 1 minus 0.14.

Notable Resumes

It’s worth highlighting several teams that have interesting resume metrics based on this new system:

Louisville is ranked 17th in Resume Quality. As mentioned previously, the Cards are 5-5 in Quad 1, with only 5 teams having more Q1 wins. However, in terms of Win Quality, they rank 14th, with their best wins coming at Pittsburgh, at SMU, and at Florida State. All of those wins were on the road, but none of those opponents rank inside the top 35 at EvanMiya.com, which is why Louisville’s Win Quality ranking isn’t higher. The Cards are 26th in Loss Quality.

Memphis has a polarizing resume, clocking in at 15th in Resume Quality. There’s a big disparity between their wins and losses. They rank 8th in the country in Win Quality, with their best wins being at Clemson, on a neutral vs Michigan State and UConn, and at home against Ole Miss. However, they have taken some bad losses, including at home to Arkansas State and away to Temple, leading to a Loss Quality ranking of 32nd.

Resume Quality shows that Auburn has the best resume in the country by a mile. They are 1st in both Win Quality and Loss Quality, with their lone loss coming at Duke, which is only counted against them by 0.15. Their best wins are against Houston, Iowa State and Purdue on a neutral, and on the road against Texas and Georgia.

I believe that UC Irvine deserves an at-large bid if the bracket were constructed today. UC Irvine is 37th in Resume Quality, which would be good enough for a 10 seed. The Anteaters are 17-3, ranking 47th in Win Quality and 22nd in Loss Quality. Yes, they don’t have an eye-popping win on their schedule against a top-50 team, but they’ve won more games than a bubble team would be expected to win against their schedule. Only sporting three losses is very impressive.

How Is This Different Than Wins Above Bubble?

The premise of Resume Quality is very similar to Wins Above Bubble, partly because I think WAB is the best existing resume metric we have. There are a few notable differences, though:

• Separating the Resume Quality number into Win Quality + Loss Quality gives a much clearer picture of a team’s work than just using one number.

• Game difficulty calculations use team ratings at EvanMiya.com, not the NET or some other site like BartTorvik or KenPom.

• An “average bubble team” is defined slightly differently. For Resume Quality, I define an average bubble team as a team that would just get into the field as the last at-large team. Right now, this means that there are 45 teams with a Resume Quality above zero, almost identical to the actual number of teams that receive at-large bids on Selection Sunday. I don't know how WAB is calculated by the NCAA, but at BartTorvik.com, there are 53 teams with a WAB score above zero as of today, which is many more than will make the tournament an at-large. This discrepancy doesn't lead to big changes in the ordering of the teams according to Resume Quality or WAB. Still, I like the interpretation of Resume Quality better: If you have a Resume Quality above 0, you have a resume good enough to make the tournament.

Today, Flagg has absolutely delivered on that promise. He’s currently the favorite to win National Player of the Year, and he’s the 2nd most valuable player in college basketball at EvanMiya.com according to Bayesian Performance Rating.

Cooper Flagg’s dominance may not surprise many people, but the stellar play of the rest of this year’s freshman class has caught the attention of many. Matt Norlander wrote a great piece highlighting the many freshmen stars in the sport today and claims that this year’s class is one of the best in the last 20 years.

I’ll take it a step further: The 2024-25 freshman class is the most impactful group we’ve seen in almost a decade. At EvanMiya.com, every player in CBB gets a rating based on their value in the sport. If you look at the top 200 players nationally, there are more freshmen (23) in that elite tier than we’ve seen in any year since the 2016-17 season:

Last year’s freshmen class had the lowest number of players in the top 200 of any year in my database, dating back to the 2012-13 season. Jumping from that uninspiring class to this year’s immense crop of talent is quite impressive, especially when you consider that there are still fifth-year players in the sport. This rule, implemented during COVID, has only made it harder for true freshmen to make a difference in the sport since.

This Year’s Top Freshmen

The table below shows the best freshmen in the sport at EvanMiya.com right now. Cooper Flagg is far and away the top player in the class, but his teammate Kon Kneuppel is right behind him. Kasparas Jakucionis comes in at 3rd, leading an Illinois team that is as fun to watch as any in the country.

Impact By Class

Last season, more than half of the best players in the sport were seniors, as highlighted by the graph below, likely the highest mark ever. This year, the number of seniors in the top 200 has dropped below 50%, but they are still the most dominant class in the sport by far.

I expect the percentage of seniors in the top 200 to drop even more next year, assuming that the rules revert to players having only four years of eligibility for next season.

#1 Johni Broome (Auburn)

Broome is hands-down the best player in college basketball this season. Like Zach Edey last year, Broome is comfortably number one in all of the advanced player metrics that matter. He’s #1 in Bayesian Performance Rating by a wide margin:

While comparing BPR across seasons is not entirely apples-to-apples, Broome’s BPR of +10.66 is better than Edey’s +10.35 last season. Broome’s per-possessions stats are arguably even more efficient than Edey’s: Broome’s Box BPR of +8.79, which estimates value from your individual stats only, also tops Edey’s mark of +8.71 in 2023-24.

Why is Broome’s statistical output so impressive? He’s contributing in every single area of play to good effect. Statistically, he’s the best rebounder in the country right now, per his nation-leading projected offensive + defensive rebounding rate of 42.3%. He also has an A+ grade in Scoring Volume, in the 99th percentile nationally. Broome is one of the elite passing bigs in the country, in the 96th percentile in playmaking score, which combines assists (89th percentile) and turnovers (60th percentile). He's virtually unstoppable when you throw in his three-point threat as well.

#5 Eric Dailey (UCLA)

Nothing from Dailey’s box score statistics jumps off the page as being elite, but his numbers are eye-popping when you dig into his advanced team impact metrics.

Eric Dailey has a team-leading Adjusted Team Efficiency Margin of +45.0, meaning that UCLA is outscoring opponents by 45 points per 100 possessions while Dailey is on the floor, adjusted for opponent strength. The next closest teammate is a full 10 points worse than Dailey, and William Kyle is more than 30 points per 100 possessions worse than Dailey.

To put that in perspective, in a 70-possession game, UCLA would outscore an average opponent by 32 points with Dailey on the floor, compared to just 9 points with Kyle playing the full 40 minutes. That’s an INSANE performance drop from Dailey to the rest of the team.

The On-Off splits (found on the Players page of the Team Breakdown tool at EvanMiya.com) tell a similar story: UCLA’s is 23 points per 100 possessions better offensively and 10 points per 100 possessions better defensively with Eric Dailey on the court compared to on the bench. His presence alone is making an astronomical impact on UCLA’s performance.

The table below shows the top pairs of players for UCLA based on how efficient the team is with each duo on the floor. Eric Dailey is present in 7 of the top 9 duos on the team. It’s hard for UCLA to find success without him being out there.

#8 Joseph Tugler (Houston)

Joseph Tugler is the best defender in the country right now on a per-possession basis. He leads the nation in Defensive Bayesian Performance Rating, ahead of Cooper Flagg, Dillon Mitchell, Cam Matthews, and Johni Broome. Though Tugler likely won’t ever be the first player you notice on Houston’s roster, he’s a huge key to their #1 ranked defense at EvanMiya.com.

Tugler’s Adjusted Team Defensive Efficiency of 72.7 is second best in the country, only behind teammate Milos Uzan. After adjusting for the quality of opposition, Houston only allows 72.7 points per 100 possessions when Tugler is on the court.

In his sophomore season, Tugler has shown some big improvement in key areas compared to last year. The chart below displays Tugler’s “big man” player skill percentiles, where he ranks 97th percentile or better in blocks, offensive rebounds, on-court defensive impact, and on-court offensive impact:

Tugler’s increased rim protection is particularly noteworthy. He averaged barely a block per game as a freshman, but this year, he’s up to 2.5 per contest. His projected block percentage of 8.9% is now in the top 1% nationally. His steady improvement across his time in college is apparent in the game-by-game career progression graph below:

#9 Ja’Kobi Gillespie (Maryland)

Ja’Kobi Gillespie has become one of the best transfers from this summer’s portal cycle, bringing his talents from Nashville to College Park. The 27th-best transfer at EvanMiya.com, he’s quickly ascended to being one of the most valuable players in the nation, now inside the top 10 in BPR.

His “guard” player skill percentiles chart at EvanMiya.com is really impressive:

Gillespie has improved in several key areas since last year, including his 3PT shooting, FT shooting, and turning the ball over. His career progression in shooting from distance is particularly noticeable. The graph below shows his game-by-game improvement in projected 3PT shooting % across his career. After being below 34% his entire freshman year (in purple), he made significant strides as a sophomore to get close to 36% (in blue) and has taken another massive leap at Maryland, shooting almost 42% this year and moving into the top 5% nationally in 3PT Shooting skill.

#12 Josh Hubbard (Mississippi State)

Josh Hubbard is one of the most unique scoring guards I’ve seen in recent memory. He’s excellent in many areas, as seen by his “guard” player skill percentiles chart below:

Hubbard has the rare ability to not turn the ball over while constantly having the rock in his hands. He’s in the 99th percentile in scoring volume, consistently putting the ball through the hoop. Hubbard also facilitates for others, in the 78th percentile in projected assist rate. And he seldom gives the ball away, in the top 5 percent nationally in projected turnover percentage.

To illustrate how rare this combination is for high-volume scorers, I pulled a list of every player this season who is in the 99th percentile or better in scoring volume while also being in the top 25% nationally in assist rate. All 21 of these players score a ton while still setting up other teammates to score as well:

Josh Hubbard is the only player in this list with a projected turnover percentage under 2%. In fact, all but one player (Kam Jones) is over 3%. Hubbard is in the 95th percentile nationally in protecting the ball, while most of this group is below the Division 1 average. It’s tough to score and facilitate to the level that Hubbard does while giving away as few turnovers as he has.

To dig into more of these metrics, head over to EvanMiya.com and sign up for a premium subscription, just $29.99 per month or $179.99 per year.

This offseason, we saw more roster turnover than ever before. Six high-major teams have entirely new rosters with zero returning players. Many of these teams were forced into roster changes because they hired new head coaches. Still, even teams with coaching continuity, like Illinois, Memphis, and Kansas State, have primarily new faces.

More than a month into the season, it’s time to quickly revisit some of these roster-building practices and see if teams with returning talent are performing better than teams with new talent.

Big Picture: Teams With Newcomers Are Not Meeting Expectations

Before discussing specific teams, let’s first examine how teams with at-large aspirations are performing relative to preseason expectations. To define expectations for a team, I use their preseason roster talent score, which measures the quality of the individual players, and their preseason team ranking at EvanMiya.com, which accounts for other factors like team history and coaching changes.

I will define a team performing better than expected based on preseason roster talent as one that has 1) a better current team ranking at EvanMiya.com than the preseason team ranking and 2) at least a 10-spot improvement in current team ranking compared to the preseason roster talent ranking. The same applies to teams performing below expectations: a worse team ranking compared to the preseason and at least a 10-spot drop in current team ranking compared to the preseason roster talent ranking.

In the table below, I have divided all preseason top 75 teams based on teams with returners playing at least 50% of the team’s minutes this season and teams who fall below this mark. The difference in performance is stark: 39% of the teams who are primarily reliant on returning players are exceeding expectations this year, while only 14% of the teams with predominantly newcomers are doing a lot better than expected. By contrast, over 50% of the newcomer-reliant teams are failing expectations, compared to just 14% for teams with a majority of their minutes coming from returning players.

Simply put, teams who can’t rely heavily on returning players are struggling to meet expectations this season based on their perceived roster talent.

Start From Scratch: Teams Who Return Minimum Players

Let’s look at the high majors who have opted to build their rosters around almost entirely newcomers. The table below shows the high majors this season with the lowest percentage of minutes played by returning players. Every team has returning players constituting less than 25% of their team’s minutes so far this year:

Of the teams on this list, which are mostly comprised of new players, only four (Kentucky, West Virginia, Illinois, and Arizona State) are exceeding expectations based on preseason roster talent, while nine (Louisville, USC, Cal, Vanderbilt, Arkansas, TCU, Washington, Kansas State, and Oklahoma State) are performing worse than expected.

Kentucky’s impressive start to the year with a brand new roster and a new head coach, Mark Pope, has undoubtedly been one of the year's biggest storylines. However, their success is an outlier compared to other high majors. The disappointing nature of many of these other teams further demonstrates how remarkable Kentucky’s immediate team chemistry has been.

Roster Continuity: Teams With Mostly Returning Players

Turning to teams that have retained a lot of returning talent, the table below shows the high majors this season with the highest percentage of minutes played by returning players. Every team has returning players constituting at least 60% of their team’s minutes so far this year:

Very few returner-heavy teams are struggling, as just Purdue, Notre Dame, and North Carolina are performing worse than expected. Meanwhile, 7 teams (Marquette, Wisconsin, Michigan State, Auburn, Florida, Penn State, and Pittsburgh) are thriving relative to their preseason roster talent.

I will revisit these trends again when the transfer portal season is upon us!