Remy Martin now has 3 team-MVP’s, continuing his hot streak in March
Game Summary:
It was KU’s defense, not its offense, which carried it for most of the game. But when Providence made a run to take the lead, the Hawks got huge buckets from Jalen Wilson and Christian Braun.
Remy Martin was the team’s MVP, producing over 23 points while giving up only 9. On an opponent-adjusted basis, his game against Providence has been his best in the season. In his last five games, he is averaging +5.72 points of value above bubble per game. He has been Remy Marchtin for sure.
Dejuan Harris deserves a shout-out here, specifically because he will be overlooked. He finished with 6 points and 2 assists while only giving up a junk bucket late. He started out each half well, only to come out to allow Remy to enter. On defense, he forced 2 turnovers and played pesky defense all night.
Other positives were Jalen Wilson and Christian Braun. Wilson racked up 14 rebounds once you factor in the dead-ball rebounds that he won possession of. He scored 16 points, most of them unassisted, on a night where KU’s two leading scorers didn’t have great offensive outings. CB, KU’s #2 scorer on the year, did add value thanks to solid defense and rebounding. Even when he doesn’t have his best game on offense, Braun plays so consistently on defense that he rarely gets outplayed. After giving up a layup on a back-cut (which gave Providence its only lead 48-47), CB added almost 4 points of value over the final 5 1/2 minutes, with no negative plays. He made his only shot, won 4 defensive rebounds, had an assist on Ochai’s dunk, and didn’t turn it over.
On the negative side, Lightfoot made some nice defensive plays but also found himself giving up size inside. He did make a basket in the second-half after a dime by Remy Martin. Ochai Agbaji struggled to get clean looks. He had a great defensive outing in the first half, but his second period was marred by lapses on jump-shooters. David McCormack struggled on both ends, and his 8 points scored was overshadowed by his 13 points allowed. He also rebounded poorly for the amount of time he played.
Last, Jalen Coleman-Lands played nearly 4 minutes and didn’t record a stat. He played solid defense to not give up a bucket.
The TEAM had a score of +6.14, which is its second game in a row below the +10 mark (KU reached that milestone in 4 consecutive games before Creighton). To get through Miami, KU will need a game score of +2 or better. What this means is that Miami is about 2 points a game better than a bubble-team.
After an inconsistent year on the defensive side of the ball, KU has produced 6 out of 7 solid defensive outings down the stretch. KU will want its offense to be much better on Sunday afternoon, but thankfully the Hawks have not had two below-bubble team offensive games in a row this season.
CB had his 10th team MVP of the season thanks to solid play on both ends
Game Summary:
With about 10 seconds to go in KU’s 79-72 win over Creighton, Christian Braun switched onto Blue Jay point guard Trey Alexander. With the game only being decided a few plays earlier, Braun mirrored Alexander’s moves as he attempted to take it toward the basket. After being cut off, Alexander pulled up from about 15′, looking to shoot a jump-shot. Braun mirrored that, too, getting an arm out to bother the would-be shot so much that Alexander decided to pass the ball at the last instance to avoid getting his shot swatted. But that proved unfruitful, as the pass did not connect and would result in a Creighton turnover. From there, all KU had to do was inbound and dribble out the clock.
This play was emblematic of Braun’s defense all game. No matter who he switched onto, he had the ability to defend that opponent. Another crucial play happened at the 3:28 mark of the second half. Creighton’s Arthur Kaluma had a game-high 24 points, but none of them were against CB. He did attempt to hit a mid-range jumper over Braun, but saw it get blocked straight over his head. In Braun’s 38 minutes of play, he only gave up 4 points on defense. Combined with an efficient offensive outing (1.29 PPP_af, 13 points, 4 assists), and he was nearly 10 points better than a bubble-level player.
Ochai Agbaji’s offense has noticeably dipped in recent weeks. But he turned up his defensive efforts on Saturday, only allowing 6 points on defense to go along with 5 rebounds, 2 steals, and 1 blocked shot. KU doesn’t win without either player.
Remy Martin was KU’s best offensive weapon against Creighton, scoring 20 points with a PPP_af of 1.24. And while he gave up 16 points, it wasn’t due to effort but due to Creighton making some deep 3’s over his short frame. This also shouldn’t be overlooked: Remy’s ability to generate offense allows guys like Braun, Agbaji, and Wilson to put more effort into making defensive stands and getting rebounds off misses on either end.
Jalen Coleman-Lands showed why he has been a better option off the bench than Joe Yesufu, first grabbing a miss and putting it back in the first-half and then hitting a second-half 3. His defense, while not great vs. Creighton, has been replacement-level for the year.
Jalen Wilson finished with 14 points and 14 rebounds, and a positive contribution despite some defensive lapses.
On the negative side: Mitch Lightfoot gave up a bad offensive rebound which resulted in a kick-out 3, David McCormack had trouble guarding the 3-point line and struggled to score like he normally does, Dejuan Harris got outplayed, K.J. Adams missed his only shot attempt, and Yesufu fouled a jump-shooter while turning it over twice.
The TEAM score of +6.83 was good enough to get the win. KU played down to Creighton’s level for most of the game, not putting away a scrappy team who was down to a 6-man rotation following key injuries. But in the end it was enough to win.
Remy Martin slams down 2 points, and earns his second team-MVP of the season
Game Summary:
Cook, Remy, cook. After earning his first team-MVP of the season in the 34th game on the schedule in the Big 12 Championship, Remy Martin got his second team-MVP right away. Scoring, assisting, stealing, and playing with a contagious energy; he was a spark that allowed KU to put away 16-seed Texas Southern early in the game. If he continues to play like this, KU is undervalued when it comes to their chances.
Other strong outings were from Christian Braun, Mitch Lightfoot, Jalen Wilson, Dejuan Harris, and David McCormack. Focusing on Mitch, the sixth-year super-senior has had positive value-performances in 5 of his last 6 games. For the season he is nearly adding more value than a bubble-player, showing he has improved tremendously.
With a TEAM score of +16.38, KU has played four games in a row where their opponent-adjusted value score was +10 or higher. This matches their longest streak of the season, which occurred in the season’s first four games. It may be tough to remember that streak now, but it was when KU was playing its best and didn’t look to have many flaws. Although that period was short-lived, injury and slumping play from some its regulars would plague much of the season, it seems to have returned at the right time.
Coach Self said that Martin and Lightfoot essentially give KU 7 starters, an assessment that the numbers show to be accurate. If anything, it is understated; Jalen Coleman-Lands has been playing well on the year, also. With the team strongly favored in its next two potential games, KU is rolling when you want to be.
Since 1985, (before 2022 Tourney) here are how seed-lines have performed in the NCAA Tournament:
There have been 36 tournaments (1985-2019, 2021). With 4 of each seed-line during each tournament, there have been 144 total teams for each seed line over this time period. This gives us a healthy sample from which to make observations.
Let’s begin by tracking overall tournament wins per seed line.
Seed
Wins
1
3.46
2
2.42
3
1.90
4
1.56
5
1.14
6
1.10
7
0.93
8
0.73
9
0.61
10
0.64
11
0.66
12
0.54
13
0.26
14
0.17
15
0.08
16
0.01
The 1-seed averages 3.46 wins per tournament, the 2-seed 2.42 and so on.
Round 64
The table below is fairly self-explanatory. FavTP is the average total points for the favorite (better seed), and the UdTP is the underdog (worse seed) average total points for each seed matchup. Point differential and winning percentage (for better seed) is then shown.
One thing to note is that only the top 3 seed lines are expected to win their games by double-digits. The 4-seed wins by 10+ less than half the time. Another notable factor is that the 5/12, 6/11, and 7/10 games are fairly close to one another in terms of point differential and favorite win percentage. The 8/9 is almost a true coin flip, but the 9 seeds have actually had slightly more success to this point.
There is a clear drop-off from the 1/16 to the 2/15 games. It is actually the starkest drop-off in terms of winning success. 1-seeds are 143-1 in Round 64 NCAA Tourney games; it is almost a bye game to the next round for the top seeded teams. To illustrate the relative performance by seed line, observe the following table. This table compares a favorite seed (seeds 1-8) against the next seed line down in terms of winning success. So we compare 1 v 2, 2 v 3, etc. and see how much more successful the better seed-line is than the next line down.
This table shows that a 1-seed is 10.1 times more likely to win its R64 matchup than a 2-seed. To be certain, 2-seeds are relatively safe to advance this round. But 1-seeds are just this much likelier to advance. As we go down the chart, we see that difference between seed lines shrinks. 5-seeds are barely better than 6-seeds when it comes to getting out of the First Round. If your team is projected to be near one of these seed lines or the other, getting that extra bump up a seed line can matter.
Round 32
When we get to this round, there become diverging possibilities with regards to which seeds play which seeds. For instance, a 1-seed can play either an 8 or a 9. A 4-seed which wins will play a 5 or a 12. This means that additional context is needed to understand how the bracket has historically played out.
Let’s start with the 1-seeds. Of the 143 top seeds which made the R32, 123 have gone on to win in this round (86%). Of the 20 which lost, interestingly 14 of these have lost to 8-seeds and only 6 have lost to 9-seeds. With the only 16-seed to reach this round falling to a 9-seed, this means that the 1/16/8/9 pod has had this many advance out of the R32 (to the S16):
1: 123 (85%)
16: 0 (0%)
8: 14 (10%)
9: 7 (5%)
Continuing on, here are how the 2/15/7/10 pods have gone:
2: 91 (63%)
15: 2 (1%)
7: 28 (19%)
10: 23 (16%)
Here are how the 3/14/6/11 pods have gone:
3: 75 (52%)
14: 2 (1%)
6: 43 (30%)
11: 24 (17%)
Here are how the 4/13/5/12 pods have gone:
4: 67 (47%)
13: 6 (4%)
5: 49 (34%)
12: 22 (15%)
The first thing that stands out is how good the 1-seeds are at getting out of the first weekend. At 85%, this is far more than the 63% of 2-seeds that make the S16 or beyond. Next, notice this effect on the 8 and 9-seeds. 10, 11, and even 12-seeds are more likely to make the S16 than the 8 or 9-seed because these worse seed lines avoid the 1-seed in the region. In fact, a 10, 11, or 12 is more likely to make the S16 than the 8 and the 9 combined.
The worse seed to make the S16 is a 15-seed, and this has been done twice. Two 14-seeds have also made the S16. We will focus on that round next.
Sweet 16
With another round in the books, the bracket possibilities can become even more complex. By this point, usually the entire country has seen its perfect bracket busted. If the bracket is chalk in a region, this is where the 1-seed faces the 4-seed and the 2-seed faces the 3-seed. Note that a 1-seed has faced a 4-seed 55 times (38.2%), and a 2-seed has faced a 3-seed 45 times in this round (31.3%). A Sweet 16 which has all four top seeds still alive should occur about 11.9% of the time. Since there are four regions (four occurrences), we expect a Sweet 16 with seeds 1,2,3,4 in any one of the same region to only happen in 2/5 tournaments. This just shows how crazy the bracket can get given the possible permutations and upset possibilities.
When we look at success, we see that 1-seeds have won 100/123 games in this round (81%). After accounting for those 1-seeds that have already lost prior to this round, we discover that 69% of 1-seeds make it through the S16. Below is the table showing how many of each seed-line makes it through, from the S16, to the E8:
1: 100 (69%)
2: 65 (45%)
3: 37 (26%)
4: 21 (15%)
5: 9 (6%)
6: 15 (10%)
7: 10 (7%)
8: 8 (6%)
9: 4 (3%)
10: 8 (6%)
11: 9 (6%)
12: 2 (1%)
No seed worse than a 12 has made it to the E8. We see the 1-seed power once again in how the results are skewed, as 5-seed is now less likely to advance than either the 6 or 7-seeds (and has the same number of E8 appearances as the 11-seed). By this point, the only seed expected to make it past the S16 more often than not is that top line. Even the 2-seed is batting less than 50%.
Elite 8
The Elite 8 round has with it additional pressure, as teams can taste a Final 4 berth. Interestingly, in this round the 1-seed does not have the best winning percentage. In fact, the 4,5, and 8-lines are all better than the 1-seeds. 5-seeds are 7-2 in this round, showing that if they can make it through, they do quite well in the later rounds.
Nevertheless, the 1-seeds get the most teams through this round due to the fact they get so many teams to this round. Of the 100 1-seeds to make the Elite 8, 59 have made the Final 4 (59%). Below are the teams to make it past the E8 and into the F4 by seed-line:
1: 59 (41%)
2: 30 (21%)
3: 17 (12%)
4: 13 (9%)
5: 7 (5%)
6: 3 (2%)
7: 3 (2%)
8: 5 (3%)
9: 1 (1%)
10: 1 (1%)
11: 5 (3%)
This handy graph helps you visualize the above:
The 1-seed is nearly twice as likely to make the Final 4 as the 2-seed. There seems to be a steady decline in F4 frequency by seed-line, until you reach the 8-seed. It drops back down, until, surprisingly, the 11-seed pops back up with 5 teams making a Final 4. A curious case of how the bracket has unfolded, to be sure. 59% of the Final 4 features teams in the top half of the bracket (seeds 1,4,5,8,9,12,13,16); 41% of the Final 4 features teams in the bottom half of the bracket (seeds 2,3,6,7,10,11,14,15). This is, once again, due to the skew of the 1-seed.
Final 4
The Final 4 round, or National Semifinals, is the first round that can feature teams of the same seed-line facing off against one another (ignoring the First Four/play-in games). This has happened in 18 out of the 72 times (25%). In 13 instances, it has been two 1-seeds. In 3 instances, it has been a matchup of 2-seeds. Once it has been between 4-seeds, and once more it has been between 5-seeds. No 3-seeds have ever faced off against each other in this round.
Using the same idea as before, we track the teams which won in this round and made it to the next. Of the 72 teams which won a Final 4 game to make it to the National Championship game, this is how the result are broken down by seed-line:
1: 36 (50%)
2: 13 (18%)
3: 11 (15%)
4: 3 (4%)
5: 3 (4%)
6: 2 (3%)
7: 1 (1%)
8: 3 (4%)
A point of clarity may be in order. These percentages add up to 100%, which is to say that exactly half of all teams to make the national championship round have been 1-seeds. But this does NOT mean that a 1-seed has a 50% chance of making the National Championship game. To get those odds, you’d have to divide each percentage above by 2. So, a 1-seed has a 25% chance of making the NC game, a 2-seed a 9% chance, etc. given the historical precedent.
The 1-seed dominates this round again, showing that in a battle of attrition (which is the NCAA Tournament’s loser-go-home system), the best teams tend to last the longest. The final NCAA tournament “round” (the National Championship game) will be the final round examined.
National Championship Game
Even though 1-seeds occasionally face each other in this round, (7 times out of 36), the 1-seed has a winning percentage of 64%, and captures 64% of all NCAA Tournaments. This means that a 1-seed is more likely to win the NC game than the rest of the seed-lines combined. When a 1-seed has faced a team of a different seed in the NC game since 1985, it is 15-5 (.750). For whatever reason, the best seed-line seems to outperform in this round. In fact, 4 of the losses that the 1-seed had against non-1-seeds were before 1998…since that year only 2-seed Villanova in 2016 has knocked off a 1-seed in the NCAA title game.
Below are the proportion of national champions by seed-line. Once again, these aren’t the odds of a team on that seed line winning the NC (you’d have to divide by 4 to get the odds of any one team winning), but rather how many national champions have that associated seed.
1: 23 (64%)
2: 5 (14%)
3: 4 (11%)
4: 1 (3%)
5: 0 (0%)
6: 1 (3%)
7: 1 (3%)
8: 1 (3%)
Once again, this is only data from 1985 on. Before the tournament expanded to 64 teams in 1985, there were lower-seeded teams like 1983’s N.C. State (6-seed) who won. But since ’85 the title has been mostly claimed by 1-seeds with some 2-seeds.
If anything, this trend is moving against underdogs. The 80’s saw multiple Cinderella runs, including 1985 Villanova (8-seed) and 1988 Kansas (6-seed). Since 2007, the only team that has won a National Championship outside a 1 or 2-seed is UConn (2011 as a 3-seed, 2014 as a 7-seed). There have been upsets and even Final 4 runs by double-digits seeds, but in the end it tends to be a 1-seed that wins it all.
538 Data
Since 2014, 538 has released March Madness models which project each team’s chances after the bracket is released and throughout the tournament. Using the pre-tournament data (but following the play-in First Four games), we will use these models to compare against the actual results. The reason to do this is to see if the historic results are close to what is predicted, or if the models think the seed-line success to follow a different distribution.
Starting with expected wins by seed, here is what 538 has projected in the 7 tournaments from 2014 – 2021 (with no tourney in 2020):
1: 3.12
2: 2.52
3: 1.83
4: 1.70
5: 1.29
6: 1.08
7: 0.86
8: 0.69
9: 0.64
10: 0.60
11: 0.59
12: 0.38
13: 0.21
14: 0.16
15: 0.06
16: 0.02
With more data, these numbers should become an even more accurate “predictor” of what seed-lines “should” win. Let’s now compare what 538 is saying seed-lines “should” win with what the wins the seed-lines have actually produced:
Seed Actual Wins Expected Wins Difference
1: 3.46 3.12 +0.34
2: 2.42 2.51 -0.09
3: 1.90 1.83 -0.07
4: 1.56 1.70 -0.14
5: 1.14 1.29 -0.15
6: 1.10 1.08 +0.02
7: 0.93 0.86 +0.07
8: 0.73 0.69 +0.04
9: 0.61 0.64 -0.03
10: 0.64 0.60 +0.04
11: 0.66 0.59 +0.07
12: 0.54 0.38 +0.16
13: 0.26 0.21 +0.05
14: 0.17 0.16 +0.01
15: 0.08 0.06 +0.02
16: 0.01 0.02 -0.01
The 1-seed has the biggest differential between “expected” wins and actual wins. 538 models have 1-seeds winning 0.34 fewer games per team per tournament, which isn’t a small number. On the flip side, 5-seeds are expected to win more often at 0.15 wins per team per tournament. 4-seeds are similar to 5-seeds. Clearly the 81% winning for 1-seeds in that Sweet 16 round is showing up in these differences.
We’ll next compare the Round of 64 games, the only games with set seed matchups (1 v 16, 2 v 15, etc.).
Matchup ActWin% ExpWin% Diff
1 v 16 99.3% 98.0% +1.3%
2 v 15 93.8% 95.3% -1.5%
3 v 14 84.7% 87.8% -3.1%
4 v 13 84.1% 78.5% +5.6%
5 v 12 64.6% 73.0% -8.4%
6 v 11 62.5% 61.8% +0.7%
7 v 10 60.4% 56.7% +3.7%
8 v 9 49.3% 51.2% -1.9%
As far as favorites go; 1-seeds, 4-seeds, 6-seeds, and 7-seeds have outperformed in the R64. For underdog seeds; 9-seeds, 12-seeds, 14-seeds, and 15-seeds have been better than expected. The biggest thing that jumps out is the 5/12 matchup, a trend many filling out their NCAA brackets have picked up on. The % difference can be misleading. We’ll now look at the expected number of wins (out of 144) for the better seed compared to what has actually occurred:
Better Seed in R64 Act Wins Exp Wins Diff
1 143 141.2 +1.8
2 135 137.3 -2.3
3 122 126.3 -4.3
4 113 121.1 -8.1
5 92 105.1 -13.1
6 90 89.0 +1.0
7 87 81.6 +5.4
8 71 73.8 -2.8
In terms of numbers, the 12-seed over the 5-seed is the biggest historic surprise. However, there is one better way to judge the performance of seeds against expectation.
Seed in R64 Multiple
1 2.887
2 0.736
3 0.775
4 0.688
5 0.655
6 1.030
7 1.168
8 0.926
9 1.080
10 0.856
11 0.971
12 1.526
13 1.453
14 1.289
15 1.358
16 0.346
What this shows is that the 1-seed has actually performed the best in the R64 compared to expectation. To best explain this, 1-seeds are expected to win 49.5 times more often than they lose in the Round of 64. In reality they’ve won 143 times more often than they’ve lost (143-1). This generates a 2.887 multiple (143/49.5). Moving down the line, 2-seeds are expected to win 20.4 times more often than they lose. In reality they’ve won 15 times more often, for a multiple of 0.736 (15/20.4).
Instead of comparing each round, we’ll move on to the Final 4 and see how seed-lines have done against expectation. Below shows the seed, actual frequency of the F4, expected frequency, and difference.
Seed ActF4% ExpF4% Diff
1 41% 39% +2%
2 21% 22% -1%
3 12% 10% +2%
4 9% 10% -1%
5 5% 6% -1%
6 2% 4% -2%
7 2% 3% -1%
8 3% 2% +1%
9 1% 1% 0%
10 1% 1% 0%
11 3% 1% +2%
12 0% 0% 0%
There isn’t a ton to say here. The numbers are fairly close to what is expected, with the 11-seed (unsurprisingly) overperforming.
Last, we’ll look at National Championships. This is similar to the F4 table above, but includes total NC’s by seed (since 1985) alongside percentage of national champions.
Seed ActNC ExpNC Diff
1 23 (64%) 18.6 (52%) +4.4
2 5 (14%) 7.8 (22%) -2.8
3 4 (12%) 2.7 (7%) +1.3
4 1 (3%) 2.9 (8%) -1.9
5 0 (0%) 1.3 (4%) -1.0
6 1 (3%) 1.0 (3%) 0.0
7 1 (3%) 0.5 (2%) +0.5
8 1 (3%) 0.4 (1%) +0.6
9 0 (0%) 0.3 (1%) -0.3
10 0 (0%) 0.2 (1%) -0.2
11 0 (0%) 0.3 (1%) -0.3
12 0 (0%) 0.0 (0%) 0.0
First, 1-seeds have clearly overachieved according to this table. They’re “expected” to win an NC about half the time, but in reality account for nearly 2/3 of all titles in the 64+ team era. Next, 4-seeds are “expected” to win more often than 3-seeds. This reveals an anomaly in the model data. Apparently since 2014, 4-seeds have graded out as slightly stronger than 3-seeds from an efficiency perspective. Over a longer time horizon, we’d expect this to revert to normal. We see that seeds below have a slight chance per the model, until you reach the 12-seed where it becomes extremely rare.
We can also look at how likely it is for a seed-line to have won over a certain time period. In other words, given the 36 tournaments since 1985, how likely is it that an 8 seed has won at least one National Championship?
It isn’t too surprising that it is very likely that at least 1-seed has won a NC since 1985. With 36 years of play, it is guaranteed that a top seed will be able to cut down the nets. This remains true for all of the top four seeds. It is somewhat of a surprise that a 5-seed hasn’t won a NC yet, since this should happen 72.9% of the time over a 36-year period. But the 7 and 8-lines make up for this.
We’ll close this out by showing how many tournaments are expected for a seed-line to win a NC. For a 1-seed, you’d need 1.93 tournaments, which reflects that a 1-seed “should” win a NC about every other year. Obviously, at some points a 1-seed will win multiple years in a row (currently at 4 in a row), and at other points it will be a few years when a 1-seed doesn’t win (they’ve been shut out in consecutive years multiple times, including 85/86, 88/89, 97/98, 03/04).
Once again, the “expected” data needs more years to smooth out. A 4-seed shouldn’t be more likely to win than a 3-seed, the same goes for a 14-seed over a 13-seed. If you combined seeds 10-16, we run into a calculation of a double-digit seed winning the National Championship once every 67 years. This may seem to be too generous to double-digit seeds, but remember that 4 11-seeds and 1 10-seed have made a Final 4. Although none of these have won a national semifinal game, UCLA came very close last season in 2021. Had they knocked off Gonzaga, they’d have been one win away from doing just this (and would have had about a 20% chance of winning). With more time, the unthinkable might happen. But it’d almost certainly be a 10, 11, or under-seeded 12 which won.
Remy Martin had his first team-MVP of the season, sparking KU to a Big 12 Tournament Championship
Game Summary:
Kansas took home the Big 12 Tournament with three great team outings. On Saturday, it was Remy Martin off the bench (12 points, 3 assists, 4 rebounds) who led the team in value. For the tournament, Ochai Agbaji was the best Jayhawk (+6.16 per game average). Mitch Lightfoot was also very good over the course of the three games (+4.64 per game).
Ochai’s emphatic dunk was part of a team-high 11th MVP game of the season.
Game Summary:
KU took the rubber match against TCU thanks to outstanding efforts from Ochai Agbaji and Mitch Lightfoot. Remy Martin had his best game in a few months, finishing with 10 points and 3 assists. Joe Yesufu got back on the positive side of the ledger with 3 points and 2 assists (with no misses or turnovers). Christian Braun struggled to finish but had a solid defensive outing to add net value.
KU will play for the Big 12 championship on Saturday.
CB earned his 9th team-MVP of the season, thanks to a balanced performance.
Game Summary:
KU won its quarterfinal match-up with solid performances from Christian Braun, Mitch Lightfoot, Ochai Agbaji, Jalen Coleman-Lands, Jalen Wilson, and Dejuan Harris. The team will play again tomorrow night against TCU.
In the Bill Self era, KU famously has as many home losses as conference championships. No team has won more than twice against Self at home, and many fans can recall each loss Self has had in the Fieldhouse from memory.
Just how good is Kansas at home? Since the 2010 season, the Hawks have complied a 197-10 record at Allen Fieldhouse. This is obviously amazing when compared to other programs, but how much of this record is due to the talent and coaching on Kansas, not the building? After all, Kansas is 77-36 (.681) in conference road games since the 2010 season. So perhaps the mystique of Allen Fieldhouse is merely just Kansas’s sustained talent alongside a home court advantage common across college basketball. If you shipped the talent Kansas has had over the years to Bloomington or Storrs or Baton Rouge, would you see similar home court records at Indiana, Connecticut, or LSU? Or if Allen Fieldhouse really is special in its ability to deliver wins, how can this be quantified?
The answer comes from predictive analytics. We’ll use Ken Pomeroy’s predictions, since they are the most famous in college basketball and tend to be the most accurate.
Pomeroy’s website contains a treasure trove of statistical information, but more than this, his system also attempts to most closely predict the final margins and winners of college basketball games. His background in meteorology has helped him forecast college basketball results with accuracy. The point is that his predictions can be used as a serious, accurate, objective standard.
One feature Pomeroy includes is home-court advantage. On his website, he notes that his is calculated by taking the last 60 home and road conference games, and finding the difference. KU’s home-court advantage is well-above average, but not close to the top (37h) at 3.8 points per game. If we were to conclude from this rating alone, we’d have to place Allen Fieldhouse not even in the top half of the Big 12, as Kansas St., Oklahoma, TCU, Iowa St., Texas Tech, and West Virginia all have better home-court advantage scores.
But I don’t see this stat as being indicative of home court advantage alone. In fact, it is just as likely to be a measure of poor road performance. The poorer a team does on the road, the better its home performances look by comparison. As mentioned above, KU has won over 2/3 of its conference road games since 2010. We don’t want to unjustly count KU’s road performances against it when assessing Allen Fieldhouse as a home court advantage.
Not Points, But Wins
Point differential and wins are two separate things. The two are interrelated, but the statement we intuitively know is true and want to show mathematically is not that Allen Fieldhouse produces large point differentials for KU (although it does do that often), but instead that it produces wins for KU.
Up to this point, we’ve established that Pomeroy makes accurate and objective predictions. Pomeroy’s system is a computer algorithm; it’s based on trends and correlations from large data sets. There are no biases; the computer doesn’t care which team name is which to next point total number. Pomeroy is using game data from all 350+ teams to make his predictions, making the model applicable across all of D-1 basketball. He also tests his predictions to ensure accuracy.
For instance, let’s say KenPom has a game where Team A is favored to win over Team B and is given an 80% chance of winning. Team A is expected to win, but if Team B won it would be a mild upset since this result would be expected to occur 1/5 of the time. Over the course of a season, or multiple seasons, we can test all games which featured one team with an estimated winning percentage around 80%. If the actual results of this collection of games turn out to be wins for the favorites about 80% of the time, then we know the prediction system is fairly well-tuned. If the results are off by some factor (perhaps these favorites won 70% or 90%), tweaks need to be made somewhere. This ability to fine-tune the model leads it to be more accurate, which makes it trustworthy. And this is in fact what Pomeroy does.
This is just the system we need to test Allen Fieldhouse. Note that in Pomeroy’s system, KU’s inherent skill level is accounted for. His predictions consider the talent of the players and level of coaching. Also, whatever amount of point value inherent to Allen Fieldhouse is also factored in. What we should be left with is a purely unbiased and accurate gauge of how much Allen Fieldhouse helps KU in terms of wins.
To test this, we will take each Kansas home game and the predicted win percentage at the time, and then compare to the results of what actually happened. Pomeroy’s site began offering probability predictions for the 2010 season, so we will begin here and go through 2022. For instance, KU’s first home game of the 2010 season was against Hofstra, in which the Jayhawks had a 98.7% chance of winning. KU did end up winning, so, it earned 1 – 0.987 = +0.013 Wins above Expectation (WAE).
As we continue this exercise for each game played over that period, we should expect the WAE to be around 0.000. Remember that the skill level of Kansas and home court advantage has already been built into the predictions. Remember also that Pomeroy tests his estimated win percentages, so we should be getting accurate prediction ratings for a team with the skill level of Kansas and a generic home court advantage. Anything substantially more or less than 0.000 needs explanation.
For KU, they have a WAE at Allen Fieldhouse since 2010 of, drum roll, +15.50. With a population of 13 seasons this is at a rate of over 1 more expected win per season. It’s difficult to stress how ridiculous this level of winning is. Instead of being 197-10, a team as good as KU has been against the same quality of opponents KU faced on the parameters of a normal home court would have an expected home record of 181-26. This is the difference between a 95.2% winning percentage and 87.7%, not a slight difference. Using Odds Ratio, Allen Fieldhouse has produced 2.77 times better results than expected.
Looking at each year individually, we see that only in 2018 (14-3 record) did KU have a negative WAE. Every other season was above expectation. 2018 was also the only season that KU lost more than 1 home game over this span.
We can also test how likely it is for KU to have an undefeated home record given its opponents. Since 2010, KU has had undefeated home seasons in the following: 2010, 2012, 2015, 2016, 2019. 5 seasons out of 12 isn’t bad; but how many would we expect KU to have?
Using similar methods, we can calculate the independent probabilities of an undefeated season for each season, and then add the results to see how many we would expect. For instance, for 2010 we’d predict that Kansas has a 38.5% chance of going undefeated. Since it did go undefeated at home that season, the 2010 team earned 1 – 0.385 = +0.615 Undefeated Seasons above Expectation (USAE). After calculating all years 2010-2022, we come up with an expected 1.749 undefeated seasons, or a +3.251 USAE. Note that even 2 undefeated home seasons would be above expectation, since over the course of 15, 16, or 17 games the math says that you should trip up at least once. But KU has managed to avoid defeat at home in five of these seasons.
Testing the Results
The first way to test these results is to run a simulation to see how frequently KU finishes with a WAE of +15.50. A quick simulator on Excel was built, and 10,000 sims were run. Here are the results, in terms of wins (out of 207 games played):
As you can see, assuming a random distribution of performances in terms of win/loss, the maximum number of wins KU achieved was 196 under 10,000 simulations. This is actually fewer than the total of home games KU actually won in this span. Looking at the 197-win mark, we can calculate how statistically rare such a feat is:
The z-score counts standard deviations, so 197 wins is 3.56 standard deviations above average. This works out to a 99.98 percentile. Out of 10,000 simulations, not once did KU reach the 197 wins at home. Given the distribution, we’d expect them to reach this win total once out of 5,453 simulations, which is a bit more common than 0 out of 10,000 but still a rare instance. KU’s success at Allen Fieldhouse has been off-the-charts.
Further Study
If anything, these numbers are undersold. Remember KU went undefeated in 2009 and 2008. If we stretched it back to the 2004 season, the first year of the Self-era, Kansas would be at 255 wins to 16 defeats, a slightly lower winning percentage but an overall larger population of games. The z-score might move even further right.
(Although we don’t have KenPom pregame win probabilities for these seasons, they can be reasonably estimated given opponent quality. It would take more work, and they wouldn’t be as precise as what a predictive model would have estimated. We will keep the (real) results for now, with the option of adding seasons 2004-2009 as an amendment so that the entire Self-era is captured).
A Free Lunch, Taking KU at Allen Fieldhouse on the Money Line
Outperforming an accurate and fine-tuned prediction model is noteworthy enough, but certainly Vegas would be tough to beat. After all, after years and years of KU winning such a high percentage of its home games, shouldn’t the betting market reflect the magic of the Phog?
It turns out the answer has been no. We won’t bury the lead on this section. You were guaranteed to win money if you bet a consistent amount money on KU to win each of its home games over any part of the past 15 seasons. Read that again. You could have started this strategy the first game of the 2008 season, or even for the Kentucky game this season that KU ended up losing. Bet a consistent amount on the Money line for KU to win outright and you will end up in the black.
KU has been 234-10 in AFH since the 2008 season (so games beginning in November 2007). Let’s set a consistent betting amount of $10. So, regardless of opponent or history, the strategy is to bet $10 on Kansas to WIN straight up on a Money Line bet.
Not all KU home games have had a Money Line, at least according to the archives. The first game in this span I could find that had a Money Line was KU’s November 2007 game against Northern Arizona, at -10000. Betting $10 would have netted you $0.10 for the KU win that game. Fast-forwarding to today, if you had kept at this strategy, you’d have $183.07, an increase 18x your initial investment.
To compare the increase of the S&P 500 over that period, if you started with $10 and let it sit in the market, you’d have about $30 today. Betting on KU to win at Allen Fieldhouse has proven to be a better long-term investment than the stock market. Anyone with a background in finance or betting markets knows how absurd this is.
Another way to look at this is to see how much the market would have to move on these games to make it efficient. In order for a better to have exactly $10 after betting $10 on each of these games, the Money Lines would have to collectively move 2.87 times what they were (so if KU was -10000, it’d have to be -28000; if KU was -275 it’d have to be -789, etc.).
Other Questions for Future Analysis
Switching gears some, it would be good to see how well other home courts do at converting wins. Clearly some teams will see better win/loss results than expected at home than other teams, this is the nature of statistics. But how well do other bluebloods, such as Kentucky and Duke perform? What about teams like Purdue; exactly how many wins has Mackey Arena been worth? Or is Hilton Magic real? Remember, since a team’s skill has already been factored into the pregame predictions, we should get apples to apples comparison regarding home arenas only, since team skill differential has been accounted for. The problem is that this data must be scraped, so it makes it time-consuming to try and build up any dataset. Maybe it would be prudent to start with a few big schools (Duke) and go from there? But there’s no way to avoid missing whatever home court has been the biggest advantage in terms of win/loss without running through every team, since that’s the very thing we’re looking for!
Conclusion
Allen Fieldhouse was worth over 15 wins over the course of the 2010-2022 seasons, propelling KU to an absurd 197-10 during this time. The chances this would occur by pure luck alone are at best 1/5000, indicating that something has happened in that building (crowd support for KU players lifting them up, opponents feeling pressure, refs being affected by the crowd, etc.) to get the team wins it otherwise wouldn’t have gotten. Whatever the reason, you can conclude this. Until proven otherwise, Allen Fieldhouse is the biggest homecourt advantage in college basketball and probably in all of sports. A 3.56 Z-score and 99.98 percentile is very unlikely to have been matched by any other arena during the past 13 seasons.
Christian Braun about to finish on an and-one. He earned his 8th team-MVP thanks to excellent defense on Saturday.
Game Summary:
Kansas won a defensive slugfest on Senior Day, and it was Christian Braun who played the best defense of anyone. He only gave up 3 points, while filling up the box score with a steal, a block, and multiple rebounds. The best offensive players for Kansas were David McCormack (1.34 PPP_af) and Jalen Wilson (1.80 PPP_af). They each got big buckets and made clutch free throws when points were limited.
Seniors Jalen Coleman-Lands and Mitch Lightfoot each had positive performances as well. Although there were only 4 points between the two of them, they each defended well. K.J. Adams had an offensive rebound and 1/2 FT to add nearly a point of value after adjustments. Neither Zach Clemence nor Chris Teahan recorded a stat in their few minutes of play.
Dejuan Harris had a neutral game characterized by excellent defensive plays (blocks, steals), some great passing (had 5 of KU’s 10 charted assists with 0 turnovers); on the other hand he only scored 3 points and did give up some easy baskets due to his slight frame. Remy Martin had good activity on offense to start, but lost his man a few times and gave up Texas’ first 5 points. Joe Yesufu didn’t do much in his few minutes of play.
Ochai Agbaji is officially slumping, as he recorded his third consecutive negative-value game. He seems to be pressing, forcing bad shots on offense while reverting to forgetful moments on D. He’s now down to 41.1% from 3 on the year. On the plus side, he is rebounding better than he did to start the year. This shows he is playing hard. Once the shots start to fall again, he will be someone who can carry this team far in the Tournament.
The TEAM score was helped by KU getting to overtime. But its still indicative of the Agbaji slump. Thankfully; Braun, Wilson, and McCormack all had solid outings to help get KU a win for the 39th consecutive Senior Day/Night game.
There are multiple ways a player can add value to his team in a basketball game. Attempting to capture these different ways quantitatively can be difficult. If you look at an advanced season box score, you’ll notice that most of the stats are listed in rates instead of totals or per game.
College Basketball Reference, 2022 Kansas Jayhawks through 30 games
So let’s take two players and compare them, Ochai Agbaji and Christian Braun. They have similar minutes played, PER, and TS%. As we continue to move left to right, we see that Och’s 3-pt. attempt rate is much higher, although CB gets to the line more (FT rate). Agbaji is credited with more points produced, CB is a better rebounder on both ends, a better assister, and also gets a higher proportion of steals and blocks. Ochai has a lower turnover percentage, and a higher usage.
These numbers are all wonderful to have, and do tell us quite a bit. They are also a mess to organize and make sense of. For instance, PER (Player Efficiency Rating) rates how efficient players are, with higher been better. It takes into account all offensive stats.
Notice we see that David McCormack is leading in this category. This may be surprising, but remember he shoots FT’s quite well and also grabs a ton of offensive rebounds. His turnover rate is also fairly low. Still, we see he is shooting worse than players like Agbaji and Braun, and has a far lower assist rate than the guards.
What if we wanted to compare D-Mac’s offensive rebounding rate (which is very high) to a guard’s assist rate? In other words, is he adding more or less value through his rebounds than Dejuan Harris is through his assist? You can’t look at the raw numbers, because they are measuring two different skills. But certainly there is a way to see this.
Represented below is the attempt to do just that. Breaking down offense, we’ve organized all offensive plays into one of three categories: Finishing, Handling, and Boarding.
Finishing – This measures how well and frequently a player scores. It takes into account his shooting percentage and volume, as well as if the play was assisted or not. It includes measurements of both production and efficiency.
Handling – This measures how well a player takes care of the ball and distributes it. It uses assists and turnovers to drive the result.
Boarding – This looks primarily at offensive rebounding, the main driver in a team getting second-chance opportunities. It also includes the occasional “save” which rescues a team’s offensive possession following a teammate’s poor pass that should have resulted in a turnover.
This chart is represented in Points Above Bubble, the value currency most frequently used on this website. This chart is showing PAB per 100 possessions, so it ignores any value related to endurance/stamina/etc. and purely rates each player on the same possession basis. There are a few interesting results:
The best offensive skill set for KU this season has been Agbaji’s ability to score (Finishing). He is shooting 42.2% from 3, and also has a 61.3% TS% (using Sports Reference’s calculations). In all, he is adding 7.29 Points above bubble over the course of 100 possessions (PAB/100) due to his ability to score.
The second-best offensive skill set for KU this season is David McCormack’s offensive rebounding. He is adding 5.17 PAB/100 just from his ability to get second and third-chance opportunities. And while some may wish to discount his rebounding ability by claiming it is due to poor shooting (“he just gets his own misses”), this doesn’t look to be too true as he has a Finishing score of +1.59. When D-Mac is out there, he is the team’s second-best offensive weapon.
The best at “Handling,” which is basically maximizing assists and minimizing turnovers, is Dejuan Harris. At +3.06, is far better at the true point guard skills than other guards. But notice even despite his strong Handling score, his lack of Finishing and Boarding makes his offensive value below bubble-player level on offense.
These numbers are not position-adjusted. So we expect to see the guards with positive Handling numbers but negative Boarding numbers, and the posts to have negative Handling numbers but positive Boarding numbers. But take a look at K.J. Adams. A solid offensive rebounder, K.J. has also done a decent job handling the ball for a 4/5 man. It is his scoring (Finishing) which has made his score so low.
In summary, different players will bring different things to the table. For KU’s offense in 2022, they are getting great scoring marks from Agbaji, Braun, and even Coleman-Lands. Harris is the best ball-handler, although the other point guards add some value here. And inside, David McCormack and K.J. Adams are beasts at winning additional opportunities. Jalen Wilson is positive in all features, being balanced at finishing, handling, and boarding.
