Kansas has finished its pre-NCAA’s schedule, and now sits at 22-10 overall following a 13-1 start to the season. The Jayhawks are on a slide, performing as a -3.05 per game team in March (4 games) and below bubble-level (-0.43) over their last 10 games, in which they have a 4-6 record. This series of poor play can be assigned solely to the offensive end, where KU is performing as a -5.19 per game team over the last 10 contests. The struggles to score have coincided with injuries to 2 key players, first Kevin McCullar (who has missed 5 1/2 games over this period) and lately to Hunter Dickinson, who was in street clothes for KU’s 20-point loss in the Big 12 Tournament. Relatedly, KU’s poor shooting has plagued the team over this stretch. In the last 10 games, KU is shooting 24.2% from 3-point range, despite being selective on its outside attempts. There’s no denying it. Kansas is slumping.
Compared to their seasons overall, there are a few players who have not performed well in recent weeks, contributing to this decline in team strength. Hunter Dickinson, who has been a +0.19 per game player in recent weeks, has a +3.90 per game season value. Johnny Furphy (-1.86 recent, -0.38 season) and Kevin McCullar (+0.63 recent, +4.01 season) are the other two who have seen a sharp decline in their games. McCullar has the excuse of being injured during this period; certainly not 100% in the games he’s been able to play.
Of the others, its not like they’ve picked up their play in the absence of Kevin and Hunter. Back-ups like Jamari McDowell, Parker Braun, Nicolas Timberlake, and certainly Elmarko Jackson have struggled all season (aside from 1-off games here-and-there). K.J. Adams, for his part, hasn’t seen a dip in his solid play in recent games (he’s been steady all season). Only Dajuan Harris has turned up his game in recent weeks (+3.05 recent, -0.20 season), but this has primarily been due to his play on the defensive end, something less noticeable and more variable. We shouldn’t overlook Juan’s good defense by any means, but his offense over the last 10 contests is only slightly above-bubble (+0.30 per game).
What is the solution to this dilemma? It’s tough to say. But one thing is that KU needs its full roster available. Even despite the relatively poor play of Kevin and Hunter recently, having them healthy enough to provide 32+ minutes is crucial for the team, as their “D” games are way better than whatever the back-ups normally provide. More minutes from Jackson, Timberlake, and McDowell do not make the team better. Role player, such as Dajuan Harris and Johnny Furphy, likewise add more value when surrounded by stars like McCullar and Dickinson. Harris needs skilled scorers to find in good spots in order to generate assist-value. Furphy needs space to make 3’s or cutting lanes, something both Kevin and Hunter can provide by their ability to draw extra defensive attention and double-teams.
There’s been much discussion about injuries and poor shooting, and note that both have been the source of KU’s woes and both are related. KU’s been worse from 3 in games without Kevin (21.3%) than in games with him (27.2%) over the last 10, not to mention how much better of a shooting team they were when Kevin was 100%. In all games McCullar has missed (6 total), KU is a -0.44 team (+5.68 with him). Further, when KU doesn’t have a 7’2 skilled post scorer they can feed as part of their half-court offense, this hurts their FG shooting from inside the arc as well. If KU gets its two best players healthy by the NCAA Tournament, this will give its role players a better chance of making shots.
In closing, KU’s best team consists of its “main four” (McCullar, Dickinson, Adams, Harris) alongside whoever is playing best of the others. It may not always look pretty, but this line-up is most effective this season. If Furphy is hitting shots, he’s a great asset to the other four. But others have provided a spark as well, and as long as their roles are limited, the damage they will do to the team is not likely to be insurmountable. Still, KU will need consistent positive-value performances from its main four to make a deep run in the Big Dance. Benching K.J. this season because of “spacing” is a very dumb idea, and is probably only broached by the types of people who never played organized basketball past the third grade.
Through the 2025 NCAA Tournament, this post examines all 40 seasons of the 64-team bracket (which began in 1985)1, specifically focusing on how seed-lines have performed in comparison to each other. Yes, the bracket has technically expanded to 68 teams, but effectively it is still a 64-team format. It’s just that there are four extra, play-in games (two between the four worst automatics, two between the four worst at-large teams) to determine the final 64 teams.
Round of 64 Results
40 tournaments, with 4 regionals in each tournament, mean there are 160 total seeds in this time. 160 1-seeds (Kansas has 16 of these2), 160 2-seeds, etc. It also means that there have been 160 First Round games featuring seeds which add up to 17 (1 v. 16, 2 v. 15, etc.). Here’s how these match-ups have turned out over the past 40 tournaments.
The 1-seeds average 84.23 points and allow 59.70, for an average difference of 24.53 points. They have gone 158-2 overall, winning 99% of games against the 16-seeds. From here, we see a drop-off for the favorite as the seeds converge until we get to the 8/9 match-up which is virtually even.
There appears to be a steady drop-off in success by favored seed until you get to the 5-seed, which doesn’t even win 2/3 of its games against the 12. The next chart compares each Round of 64 favored seed to the next seed down and produces a multiple. For instance, how much more likely is it that a 1-seed wins than a 2-seed? Or rather, how much more likely is it that a 2-seed gets upset than a 1-seed?
This shows that the biggest drop-off in terms of avoiding a First Round upset is from the 1-line to the 2-line. 1-seeds are 5.83x more likely to make it to the Second Round than 2-seeds. After that, there are smaller differences when you step down a seed-line. Because these are multiples, one can multiply down the line to compare seeds that aren’t next to each other. For instance, in order to find out how much more likely a 12/5 upset is than a 15/2, simply times 2.27 by 1.55 by 2.19. This gets 7.70. In other words, a 2-seed is over 7 times more successful at winning in the Round of 64 than a 5-seed is.
KU Focus R64
KU has earned 16 1-seeds, 7 2-seeds, 5 3-seeds, 6 4-seeds, 1 5-seed, 2 6-seeds, 1 7-seed, and 1 8-seed in the modern tournament era. This accounts for 39 out of 40 years, as the Jayhawks didn’t make the 1989 NCAA Tournament. In that time, KU has won 36 games (36-3). It’s projected record, given its seeds, is actually 34.3-4.7. So, KU has overachieved in this round. People may remember Bucknell and Bradley, but in terms of big upsets, that’s all there’s been. Don’t take for granted how good KU has been in avoiding opening round disappointments.
Round of 32 Results
The Second Round features games that can only be between these pods of seed-lines:
1/16 v. 8/9
4/13 v. 5/12
3/14 v. 6/11
2/15 v. 7/10
Of these possible matchups, only the 16 v. 8 game has never occurred. In the only two times the 16-seed defeated the 1, the 9-seed won its matchup against the 8. The winning percentage by seed-line in this round is as follows:
1-seed: 86%
2-seed: 68%
3-seed: 61%
4-seed: 61%
5-seed: 54%
6-seed: 48%
7-seed: 30%
8-seed: 21%
9-seed: 10%
10-seed: 40%
11-seed: 43%
12-seed: 38%
13-seed: 18%
14-seed: 9%
15-seed: 36%
16-seed: 0%
This trend begins reasonably enough, with the 1-seed being more successful than the 2-seed and so on. But once we get to the 10-seed, the winning percentage spikes back up. Even the 15-seed has won 36% of its Round of 32 games. Sure, it doesn’t get there that often, but it is 4-7 in this round (while the 9-seed is 8-75).
This is where the structure of the bracket has an effect on how seeds perform. 8’s and 9’s are generally better teams than the double-digit seeds below them, but have a tougher opponent as they face the 1-seed 99% of the time in this round. This causes the 8/9 winner to lose its Second Round game at such a high rate.
One way to look at the Round of 32 is to see which teams get through this round by grouping. We will look at each pod of teams and hope to gain some clarity.
1/16/8/9
1-seed: 136 (85%)
16-seed: 0 (0%)
8-seed: 16 (10%)
9-seed: 8 (5%)
4/13/5/12
4-seed: 77 (48%)
13-seed: 6 (4%)
5-seed: 55 (34%)
12-seed: 22 (14%)
3/14/6/11
3-seed: 84 (53%)
14-seed: 2 (1%)
6-seed: 47 (29%)
11-seed: 27 (17%)
2/15/7/10
2-seed: 102 (64%)
15-seed: 4 (3%)
7-seed: 29 (18%)
10-seed: 25 (16%)
Looking at all these pie-charts next to each other, we can see how much more likely it is for the 1-seed to win than for any other of the better seeds in these First Weekend pods. In terms of getting to the Second Weekend, it better for a team to be a 10 or 11-seed rather than an 8 or 9-seed.
KU Focus R32
This round has infamously been a difficult round for Kansas. In the Self-era, KU is 11-7 in the Round of 32 and 1-4 since 2019. Since 1985, KU is 23-13 in this round, and has failed to make this round 3 other times (2 R64 losses, 1 NCAAT miss). Given KU’s seeds, we’d expect KU to have been to 24.8 Sweet 16’s, indicating that KU is underperforming its seed-line through the First Weekend. Since we know that KU outperformed its First Round record by 1.7 games, we come to the calculation that KU has underperformed in the Round of 32 by 3.5 games. In other words, KU’s 23-13 record should be something like 26-10. Specifically, its multiple losses as a 1-seed (’92, ’98, ’10, ’23) and a 2-seed (’90, ’14, ’15) in this round have been major disappointments and contributed to the gross underperformance.
Sweet 16 Results
The Sweet 16 games begin the Second Weekend, with the bracket starting to winnow down as we approach the Final Four. It is when games tend to get tougher for 1-seeds (who will likely face a 4/5), and where a variety of potential match-ups can occur. The 2 vs. 3 matchup is the second-most-common (behind 1 vs. 4), but it has only happened 31.9% of the time in 160 regionals since 19853.
In order to best understand this round, we’ll consider which seeds make the Elite 8 what percent of time.
1-seed: 67%
2-seed: 45%
3-seed: 26%
4-seed: 16%
5-seed: 8%
6-seed: 11%
7-seed: 6%
8-seed: 6%
9-seed: 3%
10-seed: 6%
11-seed: 6%
12-seed: 1%
13-seed: 0%
14-seed: 0%
15-seed: 1%
16-seed: 0%
Somewhat illuminating is the fact that only the 1-seed is more likely than not to make it past the Sweet 16. And while the 2-seed is close to a 50% chance, this number drops to 1-in-4 for the 3-seed. Of the 640 seeds between 13-16 since 1985, only 1 has made the Elite 8 (2022 Saint Peter’s).
Let’s look at this Sweet 16 match-up from the perspective of the 1’s and 2-seeds. In other words, if a 1-seed makes this round, which seeds are they likely to face?
4-seed: 46%
5-seed: 37%
12-seed: 15%
13-seed: 3%
A full 82% of the time the 1-seed gets to the Sweet 16 it will face a 4 or 5-seed. For the 2-seed, here are its opponents by likelihood.
3-seed: 50%
6-seed: 31%
11-seed: 19%
14-seed: 0%
Similarly, 81% of the time it makes it through to the Sweet 16, a 2-seed faces off against a 3 or 6-seed. Only rarely will it get a double-digit seed in the Sweet 16.
KU Focus S16
KU is 16-7 in this round, which is better than its record in the Round of 32. For Bill Self at Kansas, his teams have a 9-2 record, further cementing the idea that he excels in the tournament games he has more time to prepare for. KU has won its last four in a row playing in this round. Only 2 of these wins were when KU was a 4-seed or worse (1988, 2004), and only once has KU won in this round as a seed-line underdog (1991). One of KU’s most devastating losses happened in this round as well (1997). Overall since ‘85, KU has outperformed seed-expectation by 1.3 games in Sweet 16 contests.
Elite 8 Results
There have been 160 Elite 8 games since 1985. The Elite 8 is the last round to ensure that no two teams of the same seed-line face each other. 107 of these games include a 1-seed, which is the most-likely seed (by far) to make this round. The 2-seed is also frequently at this game, having been there 72 times since 1985.
However, the 1 v. 2 match-up has only happened 51 times, or 31.9% of the time. There have been years when it didn’t occur at all (such as 2022 and 2023). Other oddities of this round include the fact the only 15-seed to make the Elite 8 faced not a 1, 4, or 5-seed but an 8-seed. Since the 14-seed has never made this round, the 1-seed has never faced it either. Here are a list of seeds that each seed-line hasn’t faced in this round that are possible (italics indicate seed has never made E8):
The 11-seed, which has made a surprising number of Elite 8’s (10), has faced the 1-seed (8 times) a 4-seed and a 9-seed once, but never the 5. The most likely match-up to occur that hasn’t yet is the 5 v. 7. 7 v. 8 is also somewhat likely to occur for the first time in the Elite 8, although it has occurred in later rounds.
Looking at Elite 8 win results is the same thing as showing Final 4 appearances, so we will include the following table:
We’re again struck by the 1-seeds’ relative dominance. While it is more likely than not the 1-seed gets upset before it makes the Final Four, over 2 in 5 1-seeds have made the National Semifinals. More 1-seeds have made the Final Four as have seeds 3 or worse.
KU Focus E8
This is another stressful round for KU fans due to recent history. Since 1985, KU is 10-6 in the Elite 8, having won its most recent two games. Self is 4-5 overall in this round, with Brown and Williams going a combined 6-1 before Self took over. Given KU’s seed-lines, KU has out-performed the Elite 8 round by 1.2 games. Even with Self’s struggles, KU has been a solid Elite 8 team overall in the modern NCAA Tourney era.
Final 4 Results
The Final Four is the first round in which seed-lines can face off against each other, something that happens with some frequency with 1-seeds and almost never with other seeds. Let’s look at who the 1-seeds face when they make the National Semis:
1-seed vs. 1-seed (15 times, or 30 total 1-seeds)
1-seed vs. 2-seed (12 times)
1-seed vs. 3-seed (7 times)
1-seed vs. 4-seed (7 times)
1-seed vs. 5-seed (2 times)
1-seed vs. 7-seed (3 times)
1-seed vs. 8-seed (1 time)
1-seed vs. 9-seed (1 time)
1-seed vs. 10-seed (1 time)
1-seed vs. 11-seed (2 times)
Counting each contest, being sure to count the 1-seed vs. 1-seed games twice, this reconciles us with the 66 total 1-seeds to make the Final Four. 1-seeds have obviously gone 15-15 against each other in these matchups. In the other 36 matchups, where a 1-seed faced a worse seed, the top seeds went 26-10 (72.2%). The last time a non-1-seed defeated a 1-seed in the Final Four (National Semifinal round) was in 2014 (Connecticut over Florida). Since that time there have been 10 straight wins in this round by the top seed when facing a worse seed.
In addition to 1-seeds playing each other 15 times; 2-seeds have faced each other 3 times, 4-seeds once, and 5-seeds once. 3-seeds have never faced each other in the National Semis.
In terms of winning percentage during the Final Four, the seeds with the best success are the 6-seeds (2-1, 67%) and the 8-seeds (4-2, 67%). The 1-seeds win 62.1% of their National Semifinal contests, in large part due to the fact they often play each other (as we showed above, 1-seeds are 72.2% winners against non-1-seeds) whereas 6 and 8-seeds don’t.
A different way to look at the Final Four round is to look at how many of each seed makes the Championship Game. Of the 78 teams to have won in the Final Four (and thus made it to Monday night), 50% of them have been 1-seeds. Here is the rest of the break-down by seed-line:
1-seeds: 41 (51%)
2-seeds: 13 (16%)
3-seeds: 11 (14%)
4-seeds: 4 (5%)
5-seeds: 4 (5%)
6-seeds: 2 (3%)
7-seeds: 1 (1%)
8-seeds: 4 (5%)
After the top 3 seeds, there isn’t much difference between the remaining seeds. The 8-seeds appear to be overrepresented, especially when you consider that no 9-seeds have made the Title Game.
An all 1-seed Final Four has occurred twice (2008, 2025), while a no-1-seed Final Four has occurred three times since 1985.
KU Focus F4
KU has made 10 Final Fours in the modern tournament era. This puts it third, behind Duke (14) and North Carolina (12). KU has gone 6-4 in this span, and it has won 4 of its last 5 National Semifinal contests. Interestingly, KU has faced only 6 different teams in this round since 1985. It has played North Carolina three times, Duke twice, Villanova twice, and Maryland, Ohio State, and Marquette once. And this is just in relation to the Final Four round. The Championship Game has seen KU face off against Duke and North Carolina as well during this time period. Given its seeds over the years, KU’s 6-4 record has put it 0.0 games against normal in this round.
National Championship Results
The National Championship game is distinct from the Final Four or National Semifinal round despite being played at the same place. Many get this confused for some reason, or ignore the magnitude of winning a Final Four game just to get to the National Championship game. There have been 40 National Championship games in the modern tournament era. Here are the teams who have won (total titles in parentheses):
Connecticut (6)
Duke (5)
North Carolina (4)
Kansas (3)
Kentucky (3)
Villanova (3)
Florida (3)
Louisville (2)
UCLA (1)
Indiana (1)
Syracuse (1)
Michigan State (1)
Michigan (1)
Arkansas (1)
Arizona (1)
Virginia (1)
Maryland (1)
Baylor (1)
UNLV (1)
What we should first note is that the 1-seeds have dominated National Championships. 1-seeds have collectively won 65% of the Titles, or 26 of 40. The other 14 titles were won by 2-seeds (13%, 5 titles), 3-seeds (10%, 4 titles), 4-seeds (5%, 2 titles), 6-seeds (3%, 1 title), 7-seeds (3%, 1 title), and 8-seeds (3%, 1 title). The 5-seed has never won.
1-seeds have faced off against each other 10 times, obviously going 10-10 in these contests. In the other contests, those of a 1-seed against a non-1-seed, the 1-seed’s record has been 16-5 (76.2%). This dominance is quite significant and helps to explain why 1-seeds appear to be overachieving their National Title numbers. However, this holds if we look at Title Games without 1-seeds.
In the 29 National Championship games which saw two different seeds face off against each other, the better seed’s record is 22-7 (75.9%). This seems to be remarkable. In the first four rounds of the entire tournament since 1985, seed-favorites have only won 71.5% of their games. In other words, seed-upsets are more common the first two weekends than in the final weekend, even though there are multiple seed-lopsided games in the earlier rounds of the tournament (1 v. 16, 2 v. 15, etc.). One would expect more 2-seeds and 3-seeds to defeat 1-seeds, or more 3-seeds and 4-seeds to defeat 2-seeds in the National Title game. But it happens relatively rarely. It was mentioned that 1-seeds have faced each other 10 times in the National Championship game since 1985, but we should add that once have 3-seeds faced each other for the Title (Michigan/Seton Hall in 1989). In 40 64-team Tournaments, never have two 2-seeds faced each other in the National Championship game. This seems almost impossible.4
The average game margin has been 8.7 points. The closest games were 1-point differences (1987, 1989), and the 2008 game was an overtime game decided by 7 points after 45 minutes. The biggest blow out was 30 points (1990). Of all rounds, the National Championship round has seen the closest end-of-game margins, so there’s an argument to be made that the best games have been in this round.
KU Focus NC
KU has played in 6 National Championship games since 1985. These games, in order, are as follows:
1988. (6) Kansas 83, (1) Oklahoma 79
1991. (3) Kansas 65, (2) Duke 72
2003. (2) Kansas 78, (3) Syracuse 81
2008. (1) Kansas 75, (1) Memphis 68
2012. (2) Kansas 59, (1) Kentucky 67
2022. (1) Kansas 72, (8) North Carolina 69
KU has gone 3-3 in the National Championship game in the 64-team era. All 6 of its games have been decided by single-digits. These games have been exciting.
If we consider KU’s seeds over the years, we’d expect them to have won 3.0 National Championship games in this span, which is right where they are. In spite of disappointing losses, “coulda, woulda, shoulda” games, and upset defeats; KU’s 3 national championships are no underachievement. In total, KU overachieves in getting past each round except the Round of 32 and Elite 8. Combined with the fact KU already gets great seeds to begin with, KU’s achievements in March Madness since 1985 have been elite.
3 NCAA Tourney MOP’s in the 64-team era.
Seeding began in 1979 for a 48-team tournament, so there is more data that could be used. However, we will stick with 1985 as the beginning of our exercise for a few reasons. One, the bye-game that the top 4-seed received into the Round of 32 did affect the bracket. If seeds 1-4 had to play First Round games in these seasons, there would have been upsets that would have reverberated throughout the tournament. Two, 1985 is a great starting point because it closely aligns with two other modern innovations. The shot-clock was first used for the 1986 Tournament, and the 3-point line was first introduced for the 1987 Tournament. Therefore, the past 39 years mostly have what we would consider the modern game. A shot-clock, 3-point line, and 64-team bracket. ↩︎
The 2018 season, in which KU officially vacated both their 1-seed and Final Four appearance, is included throughout the numbers in this post. While we don’t include these achievements when comparing KU’s status to other college basketball programs, we are keeping the 2018 results in this exercise for a few reasons. One, KU competed that season with the belief that their results were legitimate. Two, it makes things easier for our dataset. Three, the violation (money from Adidas rep to Silvio De Sousa’s guardian) is hardly a huge violation. Four, there’s a chance these vacated games get revalidated given the NCAA’s recent troubles in the court system. ↩︎
The 1 vs. 4 Sweet 16 matchup happens 37.8% of the time. ↩︎
If we accept that 2-seeds independently have a 16% chance of making the National Championship game, which is their actual total, then we’d expect 2-seeds to face off 3% of the time. In 40 seasons, we’d expect this to occur with a 65.7% chance. So, apparently it isn’t impossible for this not to occur yet. Still, we might expect 2-seeds to make the Title game more frequently than 16%, which if they did would improve their odds of having two 2-seeds face off in the NC game. Bump the 2-seed odds to 25% for a NC appearance, and there’s a 93.4% chance two 2-seeds would face each other in 40 National Title games. ↩︎
The 2024 Kansas Jayhawks finished the regular season portion of the schedule at 22-9, with a 10-8 conference record. Both of these marks are disappointments for a team which came into the season as the #1 team in the AP poll. To add injury(ies) to insult, KU’s two-best players on the season (Kevin McCullar and Hunter Dickinson) are questionable moving forward as the Jayhawks attempt to navigate the treacherous postseason waters.
Worst Case Scenario: No McCullar, No Dickinson
If KU were without the services of McCullar and Dickinson for the season, the team would have played below a bubble-team. Its other 7 rotation players average -2.13 points per 100 value. Even if you could manage the minutes to play the best players more (Adams, Harris) and the worst players less (Jackson, McDowell); the team would still struggle to be bubble-level with this 7-man squad. One glimmer of hope is that this collection of 7 players has performed better of late. In the last 10 games, these 7 players have been collectively a +2.46 per game team. While there’s no guarantee they can play this way without Hunter/Kevin (i.e. some of this value has been due to getting assists on baskets by these players), the team is actually more balanced now than it’s been anytime this season.
Possible Case: Getting 1 of them Back
Quite possibly, KU could get either Hunter or Kevin back but not the other. If Kevin’s out, this would look much like it has earlier in the season when he couldn’t go. Hunter would get a heavy diet of touches and look to score against or pass out of double teams. In the 5 games where KU played without Kevin, the Hawks averaged a +3.15 game score (+5.67 with him).
Getting Kevin back but not Hunter would be better than getting neither back, but there’s less info on how this would play out (the only time KU hasn’t had Hunter in is when he’s getting a breather, but that’s quite a big difference than not having him for a whole game or games). Kansas with Kevin but without Hunter would play smaller and likely attack defenses differently.
Best Case: Both Come Back
If both come back by the First Round of the NCAA Tournament, this KU team has a decent chance of making the second weekend. If KU can limit the minutes of Jackson, Braun, and McDowell to about 10-15 total per game, the other 185 will be filled by guys who are all better-than-bubble or trending near bubble-level. In conference play, Nick Timberlake is actually an above-bubble player (+0.02 per game). He defends adequately and has rebounded well for a wing-guard. His shooting has been a disappointment, but he limits his turnovers and doesn’t force a ton of bad shots. Johnny Furphy has dropped off in the last few games, but he is still near bubble-level for the season (-0.16 per game). Dajuan Harris has made the biggest climb in recent weeks, as he’s finally starting to defend at an all-conference level again. In total, his +1.42 per game value-score in conference play (+2.71 per game over last 10) puts him where he is adding value most nights. These three, alongside Adams/McCullar/Dickinson, make for a solid-enough team that can hang with any team on a neutral court. It may need a few breaks along the way, but this team would be worth paying attention to.
Assessing the Regular Season
KU was projected to be a +8.51 per game team, and it finished at +5.26 per game. This was a miss of 3.25 ppg, which is quite large in the grand scheme of things. At different points of the season the Jayhawks have had problems with guard-play, incoming players, depth, road games, injuries, and outside shooting. At the moment, all of these factors are contributing to what is looking like a disappointing season.
Big 12 Tournament
The Conference Tournament begins for Kansas on Wednesday, and the Jayhawks will likely face Cincinnati (could face West Virginia). As it’s looking more and more likely that both Hunter Dickinson and Kevin McCullar will miss this week to recover for the NCAA Tournament, KU is down to 7 scholarship players (of whom only 1 has produced above-bubble value this season). We will deign this group the “Scrub Squad.”
Utilizing the value-scores produced by each player, but looking at this from a per game and per 100 possession basis (while estimating minutes played), and additionally looking at the value-scores both from a season-long perspective and a recent-game perspective; we predict the per game value of each player should be around this:
Dajuan Harris +1.26
Elmarko Jackson -2.83
K.J. Adams +3.48
Nicolas Timberlake -0.70
Johnny Furphy -0.60
Parker Braun -0.83
Jamari McDowell -0.57
The TEAM score is -0.80. Given Cincinnati is playing a touch above bubble-level, but also that the game is in Kansas City, our objective prediction is that this is a 50/50 matchup. Cincy will be hungry to make a run if it wants to make the NCAA Tournament (it may need to win 3 or 4 games in the B12 Tourney). However, the Bearcats aren’t a great outside shooting team. KU has a chance to outscore them from the perimeter, something the Jayhawks haven’t done to an opponent since…the Kansas/Cincinnati game in January. If KU faces West Virginia, KU should be an 8-point favorite.
For a detailed look at KU’s excellent job of getting late-game defensive stops when the Jayhawks need them, click here.
It’s not exactly a secret that KU has a great home-court advantage. Since 2010, the Kansas Jayhawks have gone 227-12 in Allen Fieldhouse (95.0%). In conference play during these 15 seasons, KU’s record is 124-9, a 93.2% winning rate. For reference, in all 2-seed/15-seed match-ups in the NCAA Tournament since 1985, the 2-seed has a winning percentage of 92.8%. If this feels absurd, it’s because it is.
These records are a feature of three different things. First, Kansas has been very good in this span. Here is a list of accomplishments for Kansas in that span:
14/14 seasons making the Round of 32 or further (includes ’24)
11 Big 12 regular season titles
7 Big 12 tournament titles
So, a huge reason for KU’s success has been its strength. It has had consistently great teams.
The second reason KU has been so good at home is due to generic home-court advantage. In college basketball, the median home-court advantage is 2.9 points (according to KenPom)2. Teams play better at home in a quantifiable way, one that will lead them to have more wins than if they faced the same opponents in neutral or true road games. Even if KU played home games at just an average gym, by virtue of it being the home arena, KU would have some advantage.
But, once we take into account KU’s strength through the years and the generic home-court advantage, we still see that Kansas has won more games in Allen Fieldhouse than expected. For instance, since the beginning of 2010, Kentucky has had 25 losses at Rupp Arena and Duke has had 23 losses at Cameron Indoor. Given that these three programs have, for the most part, also been consistently excellent, that’s quite a gap.
The third reason KU wins so often at home is because of “THE PHOG,” an effect we classify in ALL CAPS in homage to the sign that hangs in the Fieldhouse. And we can show this effect is real.
Utilizing KenPom’s subscription services, we note that it provides pregame estimates for each game since 20103 (which is why we focus on 2010-present and not a different sampling of seasons). These pregame estimates will serve as our main data-points in order to prove that THE PHOG is indeed a magical place.
Pomeroy’s pregame estimates are devised by his algorithm and take into account each team’s relative strengths, the location of the game, and the broader data of all college basketball games to estimate the pregame win probability for each game. What makes this important is that it is neutral to any non-quantifiable factors, such as the mysterious PHOG of Allen Fieldhouse. Even though Allen Fieldhouse is special, Pomeroy’s algorithm treats it as any other gym, and because it does so, we can test to see if the Fieldhouse accounts for more wins than the Pomeroy algorithm predicted.
As we said earlier, KU has played 239 games in Allen Fieldhouse in the 15 seasons from 2010 to 2024. In each of these games, Pomeroy has made a pregame prediction. For instance, against Kansas State last night (March 5, 2024), Pomeroy’s algorithm gave KU a 84.4% chance of winning at the tip. After recording each pregame prediction in a spreadsheet, all we need to do is add up the cumulative estimated wins to see how many wins THE PHOG is worth. After all, we can confidently say that Pomeroy’s algorithm has accounted for KU’s high team strength as well as a generic home-court advantage.
Since 2010, Pomeroy’s pregame predictions have estimated that KU should have won 207.59 games, which is 19.41 games fewer than what KU actually won (227). This comes to 1.29 wins per season. THE PHOG effect is real.
To better show this, we tested to see how likely it was for KU to overachieve its home win total by 19.41 games over this period. Using the KenPom pregame predictions for each of the 239 games, we simulated these home games 10,000 times to see if we could reach 227 wins just by random chance. The results are below:
The average number of wins was 207.56, very close to the 207.59 total that was the sum of all KenPom pregame predictions. This shows that the 10,000 simulations was high enough to produce meaningful results. From here, note that the maximum number of wins produced through randomness was 223. Using the dataset of 10,000 simulations, we next show that the z-score for 227 wins is 4.09! or in the 100.00th percentile. A better way of understanding this is that we expect to have 227 wins or better only 1 time out of 46,123. This is further confirmation that THE PHOG is real and all must pay heed.
Betting Odds, A Free Lunch?
Betting on Kansas to win at home has consistently proven to be a winner. Using data going back to 2008, we estimate that if a bettor placed a $10 money-line bet on each home game4 since (276 total games), he’d have $224.81 today. This is a 20.7% annualized return. Additionally, one wouldn’t have gone negative at any time. Only twice in 15 years would this bettor have had a negative profit for that season (2011 & 2018). Betting on Kansas to win in Allen Fieldhouse has been more profitable and safer than the stock market.5
This shouldn’t be possible according to conventional theory. At some point, KU’s home-court advantage should be priced into the market. However, this strategy has only gotten more profitable in recent years. Since 2019, the average season has netted $21.24 (using $10 money line bets), more than the $13.22 average season net since 2008.
The market is likely doing the same thing as KenPom’s algorithm. It prices in the fact KU is usually the better team, and it sees a strong home-court advantage. Still, the market is trying to account for the fact that road teams do occasionally pull the upset but is still falling for seemingly-attractive prices for the road team that nevertheless aren’t equal bets due to how strong THE PHOG is.
Comparing Other Arenas
UPDATED FOR 2025
The fact Allen Fieldhouse continues to produce wins is a testament to its status as the world’s largest home court advantage.
In order to make this study more robust, we’ve begun to look at how much aura there is at other home court arenas. As this takes time to gather the data, so far only the following schools (and arenas) have been examined. These are ranked by z-score. In addition, we list Kansas and Allen Fieldhouse.
The teams selected are traditionally successful and have home courts which have reputations for being tough venues for opposing teams to win at. As these teams play different schedules and have unique strength-levels, comparing z-scores is the best way to isolate out the effects of the home courts themselves. While venues like Hilton Coliseum and Mackey Arena are quite strong (matching their reputations), none compares to Allen Fieldhouse.
The far-right column normalizes home games played in terms of 100 games. In other words, Allen Fieldhouse is responsible for 7.50 wins per 100 games KU plays there. Contrast this with Cameron Indoor, which has only earned Duke 0.28 wins per 100 games over the last 16 seasons (2010 – 2025).
Update Following 2025 Season
The Kansas Jayhawks did not have the best 2025 regular season, and ended up losing 3 home games for the first time since 2018. This tied the team’s highest number of home losses in the Bill Self era.
Overall, KU went 14-3 in Allen Fieldhouse in the 2025 season, or -0.219 games worse than KenPom expected (looking at his tip-off win percentages). So even though KU underperformed at home this season, it wasn’t by a whole lot.
Including the 2025 data and rerunning the simulations, we see the PHOG effect slip a bit but it is still very pronounced.
This shows that if you ran 10,000 random scenarios of KU basketball at Allen Fieldhouse since the 2010 season, you’d expect the team to win 237 games at most, and that only 1 time out of 17,205 would you expect KU to win the 241 total games it actually has at the Phog over this period. KU’s total Wins Above Expectation (WAE) is now 19.19, or 1.19 wins per season. So on average the PHOG is worth over 1 win per season when compared to an otherwise good college home court advantage.
Betting in 2025
Going 14-3 at home would have lost the hypothetical bettor taking KU with the money line each game. Had he bet $10 at the start of each game on the money line, he’d have lost $11.73 in the 2025 season. Now he’d still be up $213.08 since 2010 using this strategy.
Had he bet 10% of the pot each time (a strategy that would have turned $100 into $829.63 from 2010 – 2024), he’d have lost 12.3% of his incoming stake during the 2025 season (but would still be at $723.58 at season’s end if he started with $100 in the 2010 season). This would be a 13.17% annualized rate of return.
The 2020 season was cancelled a few days before Selection Sunday, and no official bracket was released. However, KU was a consensus #1 seed regardless at that point. ↩︎
KU’s generic home-court advantage on KenPom is 3.8 points, or 31st in D-1. ↩︎
KenPom users may wonder where the 2010 season’s pregame predictions are as Fan Match only goes back to 2011. However, KenPom’s win probability charts for each game go back to 2010, and we can see what the win probabilities were at tip off. ↩︎
Some of the buy-games didn’t offer a money-line option, and more recent ones are something like -100000 (meaning you’d have to lay down $10 to win $0.01) ↩︎
One difference is that you have to keep betting $10 even if you are negative, which makes this strategy inherently riskier than typical diversified stock market investments. An alternate suggestion would be to bet 10% of the portfolio (i.e. start with $100, bet $10, if you win [say $3 back], you have $103 in the portfolio and bet $10.30 for the next game). This strategy, since 2008, gets you to a $829.63 portfolio. This is an annual return of 13.6%. ↩︎
Since 2010, UNC has also played 2 home games at Carmichael Arena ↩︎
Since 2010, Gonzaga has played at least 1 home game at Spokane Arena ↩︎
Note: This post is one in a series in which we examine the NCAA Tournament in some specific detail. Terms such as “the NCAA’s,” “Tournament,” “Tourney,” “March Madness,” or “March” refer primarily to the NCAA Tournament. Today’s installment looks at the relative importance of guards in the success of NCAA Tournament teams. Initial date of publication: 2/23/2024.
The purpose of this post is to examine if guards really do win in March. There are many claims college basketball fans make that are to the effect that good guards are primarily what matters in the Tournament. Some say that Cinderella runs are fueled mostly by great guard-play, others will extend this to better-seeded teams and say that Final 4 or National Championship teams tend to have elite guards. Others will take this idea even further and claim that interior-based teams which rely on post-play are destined for early exits due to their playing style, regardless of how good they were in the regular season.
Logically this doesn’t really make sense. If a team wins during the season with elite big-men who dominate in the post, why can’t it do so in the NCAA Tournament? After all, it’s the same 40-minute game in the Tournament as it was all year. Sure, maybe there are factors which make March basketball distinct (perhaps the games are officiated slightly differently, etc.), but for the most part the game is mostly the same game it was all season. We should see all types of playing styles work in March.
At the same time, the idea that guard-play is of heightened importance in March does seem to have a hint of truth to it. Looking back to last season’s Tourney (2023 NCAA’s), the Purdue Boilermakers Zach Edeys were a dominant regular season team which earned a 1-seed only to fall in the First Round to 16-seed Fairleigh Dickinson as 23-point favorites. FDU was a guard-oriented team1 that attacked Purdue from the perimeter. Although less-remembered, the 2023 Arizona Wildcats earned a 2-seed on the backs of a strong front-line (with weaker guards) and got bounced by a veteran Princeton team. Other forward-dominant teams also underachieved, whereas strong guard-play propelled Miami and Florida Atlantic to Final 4 appearances. 5’8 point guard Markquis Nowell single-handedly lifted K-State to an Elite 8, averaging 23.5 PPG and 13.5 APG on efficient shooting during his four-game run.
These anecdotes aside, it would be better if we could test to see if a guard-based style is better than a forward or interior-based style in March. But how could we do this? Let’s first by defining the problem and then introduce some terms that will help.
Guards to Posts – A Continuum
We will conceive of how teams play as a playing Style. Teams which feature guards for higher-than-normal minutes or rely on guards for higher-than-normal production will be classified as having a more guards-based style. Teams that rely on interior players to eat minutes or produce value will have a more interior-based style. This metric will be on a continuum. Certain teams will be very guard-heavy; others will be very interior-heavy. And in the middle there will be a bunch of teams with a more-balanced attack.
Terms
In discussing Style, we are attempting to understand where a particular team falls on this guard to post continuum. Style can be used to describe teams in other ways; such as fast vs. slow tempo, man-defense vs. defense with zone-concepts, and so on, but for our purposes in this post when style is used it is referring to how guard-dominant or forward-dominant a team is.
Relatedly, the two poles of this spectrum are guard-oriented and post-oriented. We will use other terms to describe these extremes, such as guard-based or guard-led or guard-dominant. The same terms will be used for the other pole, just with post or interior or forward used instead of guard.
Other terms will be defined when we get to them, but for the rest of the blog-post, keep the above continuum in mind when considering a team’s playing style.
Problem
Teams clearly play different styles, but how do we quantify these differences? What makes a team guard-based or forward-led? In order to classify teams into playing styles, we will need to get each team’s roster of players and classify each player into a certain position. From here, we will need to find a way to appropriately weight each player’s contribution to the team’s style. For instance, a team which has a slew of walk-ons that are all guards but relies on and plays a forward-heavy rotation is better defined as forward-dominant.
Data
We used Bart Torvik to collect the data, including player positional data. Torvik has already classified each player into a position based on his own algorithmic criteria. We will use his data and trust his insight on a basis from which to build our own analysis. Getting to player position, Torvik has classified players into 8 different positions. Using the 1-5, PG-C basketball-position concept, we will assign each position into a numeric role so that a quantifiable analysis can be achieved. These roles, and corresponding numbers are below.
Pure PG (1.0)
Scoring PG (1.0)
Combo G (1.6)
Wing G (2.0)
Wing F (3.0)
Stretch 4 (3.7)
PF/C (4.3)
C (5.0)
Some of these positions are hybrid-roles, so their numeric value is in between two positions. Torvik doesn’t list anyone as a pure PF (4), but many of these types are filled in the PF/C role with some getting a Stretch 4 role. Stretch 4’s aren’t traditional 4-men, but instead a hybrid between a 4 (bigger, good defensive rebounder) and Wing F (can hit outside shots).
We limited data to include only the 68 NCAA Tournament teams. We then gathered the following information from Torvik for each player that earned 10% of minutes for his team.
Player Name
Team
Role (position)
Min%
Box Plus-Minus (BPM)
Analysis
There are two different ways that team style can be calculated. One is what we call Style-Value. Style-Value (S-V) is when a team’s style is determined by how much value its guards bring in relative to how much value its interior players bring in. Value itself is calculated through using BPM and converting it to Wins Above Replacement (WAR)2.
The other way to determine team style is by Style-Minutes. Style-Minutes (S-M) calculates style by seeing how many minutes each position plays. Guard-heavy teams will often have multiple point-guards or play 4-guard lineups. Forward-heavy teams will be more traditional in terms of minutes allotment, often having two PF/C types on the floor and/or multiple Wing F’s at times.
These two styles have some correlation to each other, but it isn’t super-strong. We will consider both S-V and S-M in our analysis.
After normalizing style, we see that these teams were the most guard-like in terms of S-V among 2023 NCAA Tournament teams:
Penn State (-2.22 z-score, 10-seed, R32)
Baylor (-1.92 z-score, 3-seed R32)
Kansas State (-1.83 z-score, 3-seed, E8)
UCLA (-1.58 z-score, 2-seed, S16)
Miami FL (-1.28 z-score, 5-seed, F4)
Compare this list to the most guard-like teams in terms of S-M:
Nevada (-2.38 z-score, 10-seed, R64)
Fairleigh Dickinson (-2.20 z-score, 16-seed, R32)
Vermont (-1.70 z-score, 15-seed, R64)
Missouri (-1.69 z-score, 7-seed, R32)
Colgate (-1.53 z-score, 15-seed, R64)
The top list shows the teams that had elite guard-play, particularly relative to their forwards. The second list shows mid and low-major teams that relied on guard-play, although it wasn’t necessarily that they had elite guards. When we deal with teams in the NCAA Tournament, you have to understand that each team has different goals. For a top seed, making the Final 4 is a reasonable goal. For double-digits seeds, the more likely goal is just an upset win. So we will use both lists, S-V and S-M.
There were different ways to analyze the relative success of guard-oriented teams. We first looked to see how strong the correlation was between style and overall success. We looked for correlation between style and relative success (to seed). There was only a tiny correlation between guard-oriented style and team success, with S-M seeing stronger correlation than S-V.
After viewing correlation, we looked to see if the most guard-oriented teams fared better than the most forward-oriented teams. We filtered out only the styles that were a SD more guard-like than the mean or a SD more forward-like than the mean. Using S-V, we saw that the most guard-like teams won +2.1 games more than expected overall and the most forward-like teams won -3.6 games than expected overall. This was an advantage of 5.7 wins in favor of the guard-heavy teams and worked out to 0.3 wins per team. This isn’t a small difference.
We did the same thing for S-M, and got a difference of +4.4 wins for the heavy-guard teams relative to the forward-heavy teams (which worked out to about 0.2 wins per team). Again these results are in the direction of supporting the conventional wisdom regarding guards and March.
Next we looked at head-to-head results. Previously it was just records overall. But we wanted to see if having a heavier guard-based style was advantageous in head-to-head contests. In the NCAA Tournament there were 67 games played, and the more guard-based S-V team won 36 of these games. For S-M, the more guard-based team won 29/67. These results may seem conflicting, but they actually will lead us to a later insight.
Looking at head-to-head contests but considering Wins Against Expectation (WAE), we see that guard-oriented teams overachieve on the whole. Teams that are more guard-oriented than their opponents won +2.07 more expected games (S-V) or +2.43 more expected games (S-M).
If we look at head-to-head contests from the perspective of the underdog, we see that guard-based underdogs have a +1.80 WAE and forward-based underdogs have a -0.86 WAE in terms of S-V. This comes to a difference of +2.66 in favor of guards-based teams. Looking at the same thing from S-M, this is +1.57 and -1.76 WAE, or a difference of +3.33 in favor of guards-based teams. This doesn’t seem insignificant. Playing a guard-oriented style seemingly helps underdog teams overachieve in March.
We will look at final margin. Using S-V, we see that the more guard-based team is better than its opponent by 0.7 points per game. Using S-M, this is nearly 2.0 points per game better for the guard-based team. It makes sense that if the guard-based team is winning more than expected it is also doing better on expected margin.
So far in head-to-head results, we’ve been considering which team was relatively more guard-oriented to its opponent. But it isn’t like teams can schedule which style on the bracket they get to play against. They can, of course, determine the make-up of their own style. So what happens if we look at the relative success for guard-based teams regardless of opponent?
There were 30 games which faced a guard-oriented team against a forward-oriented team (using S-V). Paradoxically from what we’ve seen from earlier results, the results viewed this way favor forward-based teams. When we look at how guards-based teams fare against forward-based teams, the guards-based teams underachieve, with -2.53 WAE and -1.6 points per game (forwards-based teams are the inverse).
But when we do the same exercise on Style-Minutes, we see that guards-based teams overachieve by +6.05 WAE and are +5.1 points per game better than expected. This shows the importance of defining style. If we say a team is guards-based by looking at the relative value that team’s guards provide, it can give a whole different answer than if we say a team is guards-based by looking at its minutes distribution by position.
Making Some Conclusions
So where does this leave us? Is style important? Do guards win in March? And if so, how can this be used by coaches to gain competitive advantage?
Based on evidence so far, we would conclude not that interior-based styles are doomed to fail, but that guards-based styles are more likely to overachieve in March Madness. While there are weak correlations seen when we regress a team’s style against its success in the 2023 NCAA Tournament, when we filter out the most guard-heavy and forward-heavy teams, there is a clear bias in favor of the guard-heavy teams (relative to pre-Tournament expectation).
Likewise, when we view head-to-head results, the team that is more guard-oriented outperforms the team that is more forward-oriented on the whole. If we look at guard-based teams in head-to-head results (regardless of opponent-style), there is a contradictory answer based on if we base style upon value (Style-Value) or minutes (Style-Minutes). In all assessments, the Style-Minutes view showed stronger benefit for guard-based teams.
Overachievement is the key word. Guards-based styles displayed a better chance of winning when looking at pre-game or pre-Tournament expectation.
However, it isn’t clear how much coaches can use this to their advantage. Particularly because they do so already. Low and mid-major teams looking to pull the upset off are already more likely to be more guard-oriented than their favored opponent. In fact, of all the correlations we ran, the strongest one we found was between a team’s pre-Tournament computer strength and its style as a forward-based team. The best teams heading into the 2023 NCAA Tournament were more likely to be interior-based when compared to the average Tourney team.
But if low to mid-major teams gain a March advantage by playing guard-based styles, can this be neutralized by favorites looking to stave off an upset by playing small themselves? This is tough to say. In one sense, coaches of top teams are best to stick with the style that got them to a good seed in the NCAA’s, as they are still the favorite against the guard-based underdog. A favorite changing its playing style to become more guard-focused might help in some areas but hurt in far more areas (such as dominating the glass and paint). This trade-off might not be worth it.
Additionally, and this hasn’t been analyzed, but favorites might already be changing up their styles during March Madness, and this is partially to blame. Perhaps forward-based teams are going away from what got them to the Big Dance once things get tight in a early-round game, and this is being reflected in the data! We just don’t know.
We’ll close this section with examining the biggest upset in the 2023 Tourney. Purdue’s loss was blamed on its guards, and it strengthened the claim of those who say guards win in March. At the same time, this ignores how well Zach Edey played. Edey had 21 points, 15 rebounds, and 3 blocks. He drew fouls and made FDU work on the defensive end. While it wasn’t enough, had Purdue escaped the upset it would have been due to the team’s interior play. The point is that Purdue wouldn’t have benefited from changing up the style of play that earned it a 1-seed. Rather, it just needed its guards to not choke. The guards/wings went 5-26 from 3. Purdue had 16 turnovers but forced only 9. It lost the game on the perimeter for sure. But if Edey didn’t play, it wouldn’t have had the big advantage inside either, which would have made things worse.
More to Come
With the 2024 tournament coming up, it will be interesting to see if these patterns hold and if this information can be used in filling out a bracket. Picking guard-based top seeds to make deeper runs (using S-V) or double-digit upsets (using S-M) might be a successful strategy. We’ll fill out a few brackets using these principles and see how they do.
Beyond this, we will want to see if this is consistent across tournaments. 2023 had quite a few upsets, so this could have been the reason guards-based teams overachieved. Maybe 2019 (all but one S16 team was a 5-seed or better) or 2008 (all Final 4 teams were 1-seeds) will show the reverse. Maybe upsets or deep-runs are more memorable when a team is led by its guards, and this clouds our view on how style plays into Tournament success.
FDU was certainly guard-oriented from a minutes-played perspective, but in terms of player-value, FDU was not a team which featured even average guard-play during the 2023 season. FDU wasn’t good at all during the 2023 season, and entered the NCAA Tournament somewhere in the mid-200’s in computer rankings. If anything, it could be seen as having better forwards than guards as its interior players were less-bad than its guards. ↩︎
WAR multiplies BPM by Min% by a set # of games by a multiple to estimate a player’s contribution to team wins. ↩︎
After Jalen Wilson departed for the NBA and KU picked up 7’1 center Hunter Dickinson in the portal, K.J. Adams shifted his position from the 5-spot to the 4-spot in KU’s starting line-up. This shift has been seen on both ends of the floor. Offensively, he’s posting up less and making more plays on the perimeter as a creator (where he’s third on the team in assists). And while he doesn’t have what one would call “range,” he’s been making some mid-range jumpers and push-shots with Hunter underneath to rebound, which is what a traditional 4-man does when he is playing with a true 5-man.
But it is K.J. Adams’ defense which has been the most-valuable part of his game with his switch to the 4-spot. Last season, as someone primarily guarding the opposing 5-man at a height and length disadvantage, Adams was basically a neutral-value defender, finishing with a Per100 points-against bubble of -0.22. He gave up 12.7 points per 60 possessions (worst among the starters) as he had to battle inside. (Adams was able to benefit by having a slower player on him on the offensive end, so he did finish as a positive-value player overall (+0.49 per game)). In 2024, Adams is KU’s strongest defender, with a +5.01 Per100 value. And while his rebounding hasn’t really improved, he is giving up only 7.7 points per 60 possessions now that he’s guarding mostly wings and forwards. This defensive stinginess not only leads the team, but is close to the range of what KU’s best defenders have allowed over the past decades.
Since 2018, we have tracked defensive value at detailed-enough level to estimate which players were the best at not allowing points to be scored. This is seven seasons of data, including 2024. In this time, only three times has a player been a stingier defender than K.J. Adams this season. Marcus Garrett (2020 and 2021) accounts for two of these seasons. The other, which you wouldn’t probably guess correctly if you had 20 chances, was Isaiah Moss. While Moss didn’t rebound or force turnovers, he stayed close to his man and didn’t overhelp. Overhelping, particularly off wing shooters, is where many teams lose defensive value. Completely leaving one’s man should only be done to prevent an uncontested layup or wide-open 3 by a good shooter. Moss wasn’t overly athletic, but he had a high BBIQ as a veteran defender.
But back to K.J. K.J.’s value on the defensive end is estimated to be +5.01 points above that of a bubble-level player Per100 possessions. If we look at this from a per game ratio, then Adams is producing an estimated +2.91 points of defensive value above a KU-caliber replacement. According to current information, this would put him as having the 7th-best defensive season in the last 32 seasons (there have been 453 player-seasons in that time. K.J.’s 7th of 453).
That’s rather incredible. Whenever K.J.’s defense is talked about by fans, it’s often derided due to him not getting enough rebounds or having “short arms.” It’s true he doesn’t add value through rebounds or forcing turnovers. But he does an excellent job of not giving up easy scoring opportunities. He uses his strength to guard bigger players and his quickness to stay in front of guards. This versatility means he can easily switch positions 1-4 (and as we saw last season, can also play some against 5’s without being dominated). In turn, this forces teams to over-pass or take a less-than-ideal shot on certain possessions. All of this shows up in the numbers when we chart each defensive possession.
Note that these value estimates aren’t an exact science. Other sources don’t see him as having this much defensive value. Hoop-explorer doesn’t see much difference (although it relies on On/Off data which is very noisy and probably not wholly accurate). Evan Miya, which uses On/Off data and attempts to normalize the relative strengths of the other 9 players on the floor, has him as KU’s fourth-best defender. Without being overly critical, this seems to discount the on/off method for assessing player value. It isn’t K.J.’s fault if another teammate blows a defensive assignment when he’s in the game. While, with enough of a sample size, this “noise” would eventually even itself out (theoretically leaving you with just the player’s impact), it doesn’t seem to be the case unless there is a much larger sampling of minutes played. While K.J. has played 840 minutes so far, this only leaves 165 which he is on the bench. On/Off here isn’t really helping.
In 25 games this season, K.J. has posted positive defensive-scores in 21 of them. Teams aren’t attacking him much (like they would last season if they had a capable big). This means he’s doing his part. KU’s had defensive lapses, particularly among ball-screen coverage, but this is mostly due to Hunter being out of position (KU would do better to drop him rather than have him hard-hedge) or a separate wing/guard not being good at rotating over. K.J.’s fewest minutes in a game came against Chaminade, after he arrived following the death of his mom and didn’t start. He wasn’t expected to play, but did get on the court for 26 minutes and produced a small positive amount of defensive value. Self clearly wanted him on the court despite the opponent being a Division II team. This is further evidence that his defense warrants playing time, and is thus valuable.
All KU fans agree that Hunter Dickinson is a skilled and valuable offensive player. Some argue that Dickinson’s defensive deficiencies (they take for granted he’s a poor defender) severely undercut his offensive value. This post will examine Hunter’s defense by comparing him to other players on this season’s roster as well as KU centers in seasons past.
A player can be a plus defender in one of two broad ways. He can either do so by limiting the amount of points his man scores or he can be a good defender by getting the ball back for his team (such as rebounding). Note that these are related. A defender getting a rebound is also preventing his man from scoring on a putback attempt. The very best defenders are good at both aspects–they limit their opponents’ good looks and they get the ball back for their team.
Applying this to Hunter in particular, we want to examine how often he is most responsible for the other team scoring (compared to others) and how often he wins possession back through a block, rebound, steal, or forced turnover (compared to others).
Using the Charting methods, we calculate that Hunter has allowed 276 points in 23 games this season (12.0 ppg). When we convert this to a 60-possession basis, this comes to 12.9 per 60. This conversion is done to compare Hunter to other players.
Next, we can count the number of stops KU gets because Hunter gets the ball back through either a block, rebound, steal, or forced turnover. As rebounds are most common way a defense gets the ball back after a stop, Hunter’s importance is mostly by being a great rebounder. He averages 10.5 possessions “won” over 60 defensive possessions.
For the 2024 team, we can consider this table.
Looking at the Points Against Per 60 column, this shows that Hunter (allowing 12.9 per 60) is worse than the other starters save Furphy (13.0 per 60). However, he is better than his backup Parker Braun (15.4). Additionally, only K.J. Adams (7.9) is significantly better than Hunter at disallowing points. Harris and McCullar have been in the middle-of-the-pack.
If we move to the far-right column, we see possession winners per 60. In this case, Hunter is clear-and-away the best at getting the ball back following a miss or through a forced turnover.
When some think of defense, they mostly consider the first aspect of this, or on-ball defense. And this aspect is very important. Not getting beat, not fouling, closing out on shooters, forcing a player to pick up his dribble and pass, etc. are all ways a defender can make it more difficult for the opponent to score. But the second aspect, or coming away with the ball, is also important. First-shot defense that doesn’t win the ball back will lead to second and third chances that can allow points. A player who might not be a great first-shot defender can still add value if he prevents opponents from getting second-chances. So, we need a way to combine these elements of defense into one number that estimates player value.
Using theory regarding the relationship to points and possession in basketball, we can calculate how valuable these defensive metrics are. We will now include the defensive value metric in the final column of the table.
This final column estimates, in points per game, how valuable a certain defender is when compared to a replacement (bubble-level) player. Due to his stingy on ball-defense, K.J. has graded out as KU’s best overall defender. But the second-best defender has been Hunter Dickinson–not because he is always great at disallowing points–but because he limits teams’ second-chances. Imagine KU without Hunter on the floor. Teams would get many more baskets through second-chance opportunities.
This is common among bigs in today’s game. The prevailing offensive strategy involves getting opposing 5-men away from the basket (to clear driving lanes and cutting angles). As Hunter is involved in ball-screen defense and plays on the perimeter at times while on defense, he is often put in situations where he can find himself out of position. This leads to breakdowns, rotations, and open shots for opponents. But this isn’t always what happens. Other times he defends fine, and the possession ends with a missed shot that he has a great chance of rebounding due to his height and good rebounding technique.
We can break down the defensive value provided by 2024’s roster further, using 3 categories and a Per 100 possession basis. These three categories are: Stinginess (or points allowed), Pressure (forced turnovers/steals), and Boards (rebounds).
Hunter’s Stinginess score is below-bubble, but at -0.65 points we estimate that it only costs KU less than a point per 100 possessions. His Pressure (+0.34) and Boards (+2.93) make up for it, leading him to be a +2.63 player over 100 possessions. This is nearly 4 points better than his backup, Parker Braun has been. He is also better defensively than all other players save K.J. Far from being a liability, Hunt’s been a valuable asset for the 2024 defense.
We can also compare Hunter to 5-men of past seasons. Below is a table of all players labeled center who’ve played 10% of minutes on the season or more since 2018. This includes 9 player-seasons.
Here we see a pattern. 8 of the 9 centers on this list have negative Stinginess value, showing that they are allowing points more often than the average defender. This makes sense. Teams are attacking the slowness of the big men to find an advantage that leads to points. However, once we move right, we see that the 5-men can Pressure and Board better than average. So while your guards are likely to be better at preventing open shots, they struggle to get the ball back. Hunter’s rebounding better than any recent center not named Udoka. Additionally, his overall defense is 3rd best out of 9 centers since 2018, again only behind 2019 and 2020 Azubuike.
We can expand this more, just with less detail. Since 1993, there have been 192 players who have played at least 40% of the team’s minutes during the season. Of this 192, only 45 have had a per game defensive value as good as Hunter’s having this year. He’s objectively around the 76th percentile of defensive players in recent Kansas history. Ignore people who don’t know how to track defensive value and fail to consider all relevant factors of defense.
In addition to tracking individual defense, we can confirm these general findings by looking at proxy metrics. One such metric would be points in the paint. This season, for reference, KU’s offense is averaging 40.1 points per game from paint scoring. That’s what a good offense does. A large chunk of this is from Hunter (although K.J. and Kevin are also doing well at the rim). So let’s use KU’s offense as a comparison to KU’s defense. The thought would be that if Hunter’s offensive-value is being wiped away by poor defense, then KU will be allowing nearly as many points in the paint as it is scoring.
This isn’t the case at all. KU is only averaging 25.8 points per game in the paint. This is a difference of 14.3 points, which is more than the overall point difference (10.9). KU is winning points in the paint but losing in all other areas (that is FT’s + 3’s + 2’s outside the lane). This is just further confirmation of Hunter’s value.
KU’s 24-point win against a hapless Oklahoma State team last night boosted its KenPom rating to +22.27, which is good enough to be 14th overall as of January, 16. There’s been some debate about the accuracy of KenPom’s system, as it appears to reward teams more for blowing out bad teams than beating good ones in close games.
Without getting into a longwinded opinion about what KenPom is “about,” I wanted to explore what I call a team’s “Game Score” and see if, in theory, a good team could improve its KenPom rating by scheduling easier.
First, let’s discuss Game Score. Game Score is the attempt to quantify how good a team’s game is by adjusting the final margin depending on the opponent and location. That is, a 10-point home win against a team ranked 250th would be much less impressive than a 10-point road win against a team ranked 25th. Game Scores tie to KenPom’s ratings in a few ways. One, they use KenPom’s Adjusted Efficiency numbers to determine opponent strength. Two, they are linear or rely on a linear system to compare game performances, just like KenPom. If a team is projected to win by 10 and it wins by 15, the team’s KenPom rating will climb. However, the hypothesis is that a good team can have a higher average Game Score (and in turn elevated KenPom rating) by playing an easier schedule, because it is easier to “juice” a winning margin in games already in hand. When playing an overmatched opponent, leaving in the starters just a few extra minutes can be the difference between say, a 25-point win or a 30-point win. The Game Score will increase by 5 points as well, as the opponent adjustment is already baked-in.
To test this, we will use KU games over the past few seasons. Kansas has been consistently good over this period and is a good team to use. My theory is that there will be an inverse relationship between Game Score and opponent quality. Remember, in theory, opponent quality should neutralize the effect of the actual margin, making Game Score independent of opponent quality. Does reality show this to be the case?
KU has played 229 official games since the beginning of the 2018 season (November 2017). These games have been played against a variety of opponents, some good, some not. Will it be easier to produce higher Game Scores against worse opponents than better ones?
Results
There was a small inverse correlation between opponent quality (opponent’s KenPom rating) and Game Score. While this might seem insignificant at first, it can make a difference when we look at the results.
Since 2018, KU is averaging a +8.03 Game Score. This just means that KU is beating its opponents by an average of about 8 points more than a typical bubble-team would. When we break it down into buckets by quality of opponent, we see a clear trend. Against the best teams in the country, those with a +20.00 KenPom rating or higher (roughly your top 15 to 20 teams, depending on year), Kansas is only producing a Game Score of +6.07. Against decent to okay teams, those with a KenPom rating from 0.00 to +20.00 (corresponding to ~ 20 – 180), Kansas is producing a Game Score of +7.72. The best Game Scores are those against inferior opponents, those with a negative KenPom rating or who are worse than the average D1 team. KU is, on average, +13.82 in these games.
What this shows is that, over the past 7 seasons, KU is more likely to climb in KenPom when it plays games against easier opponents. As KenPom is trying to account for opponent strength already, it does give strong credence to the idea that teams who play a weaker schedule are juicing their KenPom scores by doing so. This would apply to teams like Gonzaga in past seasons, but also to the non-conference schedules of Houston and BYU in 2024.
What we’ll next show is the expected KenPom ratings/rankings for a hypothetical Kansas team that is as strong as the average KU team since 2018.
The first column looks at the average opponent KenPom rating, and then projects this hypothetical KU team’s KenPom rating. The next column is the KenPom ranking for additional context. We see that as the average opponent score climbs, the KenPom rating/ranking for hypothetical Kansas worsens. A schedule of 0.00 opponents would allow Kansas to juice its score all the way to a #4 ranking this season. If it played a schedule against top-tier opponents, like it will in stretches of conference play, this ranking drops out of the top 10.
An additional level of data relates to the 374 games played since the 2014 season. KenPom provides pregame point differential estimates as well as winning percentage estimates (effectively odds). With this info, we can make the prediction that KU will “overperform” in easier games compared to games where the margin is expected to be smaller. In the 90 games where KU had a 90% chance or greater of winning, that is the easier games, KU has outperformed its expected KenPom margin by +2.41 points per game. In the 284 other games, the games expected to be closer, KU is outperforming KenPom expectation by only +0.49 points. This accounts for a difference of 1.92 points per game, or roughly 2.74 AdjEm on KP. Again, this is further confirmation that good teams can juice their KenPom ranking by playing an easier schedule.
Relevance to the 2024 season
This is relevant to Houston, a team which has a non-conference average opponent rating of -1.21 (215th), as well as BYU, a team which has a non-conference average opponent rating of -5.01 (307th). For Houston, if we look at its 6 games against Top-100 opponents, we note that it has performed where a team with a KenPom score of +27.39 would, or nearly 5 points lower than its current (juiced) +32.31 rating. Only looking at conference-play against Top-100 opponents, a small sample size for sure, its comparable KenPom score is +21.82.
BYU’s 7-game slate against Top-100 opponents has it 4-3, with an estimated KenPom rating of +20.14. This is the difference between 9th or 20th. Again, we’re seeing evidence that a team’s KenPom rating is aided by blowing out inferior opponents. For both Houston and BYU, their overall schedule strengths will get better, meaning that what KenPom says about them in March will include more data against strong opponents. As both are good teams, one shouldn’t write either of them off. But what Purdue (best overall schedule) or Wisconsin (surprise team around BYU’s ranking yet has a solid SOS) have done so far should give them the nod ahead of teams similar to them in the KenPom totem.
We can use Bart Torvik’s filter tool to explore this phenomenon more. While Torvik has different inputs than Pomeroy, he is attempting to do the same thing. If we only look at opponents that are Quad 2 or better (using NET), we see that Purdue is best (and better than Houston by almost 7 points per 100 possessions. Houston is slightly better than Purdue accounting for all games). Similarly, BYU is 26th best when playing top opponents, as opposed to 7th when all games are included.
Determining the Best Teams
So, who is the best team in the country? How much should advanced metrics be considered? Keen observers will note that, in criticizing KenPom’s rating for Houston or BYU, I still used KenPom numbers. That highlights the inherit difficult of any ranking system. Rankings are relative, they depend on how good others are. KenPom’s value is that the system provides a fairly accurate look at each team’s relative strength compared to any other of D1’s 362 teams. If a system makes adjustments to remove Houston or BYU’s blowout wins, it underestimates their strength relative to a team outside its Top-150. So for any advanced rating system to be most accurate, not just accurate among a certain range of teams, it needs to include all game data.
For the purposes of college basketball, the only important rankings are those of the selection committee preparing to set the field for March Madness. The selection committee gets criticized, sometimes rightly, but those who want to do away with the committee and institute a computer-based ranking system (such as WAB) are completely misguided1. Even the most well-known advanced rating system has noticeable flaws. The committee’s focus on strength of schedule (dissuading teams from scheduling cupcakes in the non-conference to artificially boost their overall record) appears to be a good thing. It looking at a team’s best wins or worst losses will also filter out noise, allowing a better comparison between teams near each other on the seedline.
Summary
Teams have an easier time improving their advanced metrics by playing a softer schedule that allows for blowout wins. Teams that’ve achieved high rankings by playing a strong schedule should be seen with less skepticism than teams that have mostly blown-out lesser opponents. Still, let’s not be overly critical toward systems like KenPom, which are trying to rank all teams (not just the top 50 or whatever).
The only real ranking that matters is the NCAA Committee’s “S-curve” on Selection Sunday. While the committee does get things wrong, it has avoided errors by not adopting a metric-dependent system like some of its detractors demand.
1- Gasaway states the following in his appeal to do away with the Committee: “Whether your preferred metric is wins above bubble (WAB), strength of record (SOR), or something similar, any team’s record can be evaluated according to how difficult it would be for other teams to achieve the same result against the same opponents in the same venues. Wins are the key metric, as they should be, and teams can excel whether their schedules turn out to be grueling or unexpectedly soft.” (Links in original).
His assumption is faulty. We cannot confidently say that WAB or SOR are accurate enough when its been shown that the advanced metrics themselves understate how easy it is for a good team to blowout a bad one. This isn’t to say that WAB or SOR cannot be tools that an informed committee uses, but a committee-less NCAA Tournament selection process that only looks at a computer metric would reward teams that dominate against a light schedule and punish teams that take some lumps in tough conferences but still finish around 19-12.
Wins Above Replacement (WAR) and PAB (Points Against Bubble) have similar concepts and therefore the lists will be similar. Both are attempting to rank the careers of all KU players since 1993 to see which Jayhawks have been the most valuable for the program. The differences stem from the benchmark player upon which value is judged. In WAR, the “replacement player” is basically an average D1 player whereas in PAB, the “bubble player” is a player whose value makes a bubble team no more or less likely to make the tournament. In other words, a bubble player is a much higher standard than a replacement player.
So why the two lists? Which list is more valid?
At the top of the list there isn’t a ton of difference as guys like Raef LaFrentz easily clear the bubble level just as they clear the replacement level. It is further down the list where this distinction matters. Take Jeff Boschee for instance. Boschee was a four-year starter who is currently 57th in career WAR but 168th in PAB. Boschee was better than a replacement player but not better than a bubble-level player for the majority of his career1.
These are widely different rankings, leading to the follow-up question, which list is better? Well, it depends. Being a better-than-bubble player is crucial for KU to approach the success demanded of it. So we would normally lean with PAB. However, sometimes players who are not quite at bubble-level (but still solid enough to be above replacement level) are required to play heavy minutes due to roster issues or coaching decisions, and it wouldn’t be fully accurate to list these players below low-minutes-played players because of this. Boschee serves as a great example. He was certainly a below-KU-average guard given the minutes he played, particularly as an underclassman. But he was still valuable because he was better than the alternatives at his position.2 This is where WAR can be a better metric.
There is a third way to measure that is present, namely Per100 (possession) against bubble value. This is the third column on the second table. This shows the value the player provided while on the court, effectively ignoring playing time. Be aware of the limitations of this measurement. One, if we sorted by this list we’d include walk-ons at the top who graded out very well in very small, unimportant minutes. Two, durability is an important feature inherent to both WAR and PAB but ignored in Per100 value. A player who generates high value per possession but does not stay healthy throughout the year may not help the team as much as someone who can play a full schedule and heavy minutes at a lower value. A four-year player with solid value (like Marcus Garrett) will provide more value to KU than an elite one-and-done (like Joel Embiid).
Both WAR and PAB are useful lists but cannot be considered in isolation. If we are looking toward the top of the list, PAB is the better measurement. It estimates how many points a player gave to Kansas above that of a bubble-player over his career. Toward the middle and bottom of the list, WAR becomes the better mark. Clearly role players with sub-bubble (but above-replacement) value are preferable to the walk-ons. Lastly, Per100 can add some more context but we must be careful to not over-emphasize this measure as it is total value, not per possession value, that is most important when determining this ranking.
Boschee’s only above-bubble season was his senior year in 2002. ↩︎
Marlon London and Jelani Janisse are examples of other backcourt players on the roster in the Boschee-era. KU’s backcourt talent depth was quite thin during Boschee’s first three seasons at Kansas. ↩︎
Charting the Hawks 