KU Tournament Paths in the Self-era

The NCAA Tournament also goes by the moniker “March Madness,” reflecting the unpredictability of outcomes. Kansas has been no stranger to this madness, as it has been bounced early on multiple occasions as a highly-favored team but also won the 1988 NCAA Tourney as a 6-seed.

Each path to a possible title is unique. Some years, the bracket can open up for a team as upsets clear the way for a better chance at success. Other years, the bracket goes chalk and difficult opponents are faced. We will explore how varied KU’s possible paths have been down below.

Also at play is the idea of the possible opponent. Had KU defeated Arkansas this year in the R32, its next game would have been against Connecticut in the S16. For fans of any team, viewing potential matchups down the line is commonplace after the brackets are released on Selection Sunday. But after the team loses, the would-be matchups become less intriguing. For this exercise, we will be examining KU’s would-be matchups had they advanced.

We will restrict our study to the Bill Self era. This gives us 19 seasons worth of Tournament data (Self has been the coach at KU for 20 seasons, but 2020 is excluded obviously). For data, we will use KenPom’s final results of that season to serve as a team’s relative strength. We will then match up Kansas against its actual or hypothetical opponents for any given round. For years where KU makes the championship game, these opponents will all be actual opponents. For other years, some opponents will be those KU would have faced had it kept winning. Then, using KenPom’s numbers, KU will be compared to said opponent to get an estimated points spread. Next, these point spreads will be converted to winning percentage estimates using historic data from Team Rankings and a smoothing process. For instance, if a team is 7.7-point favorites (per KenPom), then that team wins the games an estimated 78.0% of the time (using the smoothed historic numbers). It is this 78.0% number that gets used to estimate Final 4 and National Championship odds.

We’ll start with Final 4 odds. Below is a table which calculates KU’s odds of making the Final 4 given the path that occurred.

The seasons with actual F4 appearances are highlighted in green. The average F4 chance for KU in the Self-era is 23.1%, with conditional formatting being set as Red for 1/10 chance, Yellow for 1/4 chance, and Green for a 1/2 chance. There are two factors at play. First is the draw, how the bracket turns out. More upsets in your region means an easier path. The second factor is the strength of the team, and this also turns out to be important. For 2008, KU’s biggest break was in the S16, when it faced 12-seed Villanova. For a S16 game, it was a much weaker opponent than normal. The E8 game, against 10-seed Davidson, was not really a break. Despite the double-digit seed, Davidson was about as strong as the average E8 opponent or potential opponent KU faces/could face.

KU took advantage of 2 of its better chances to make the Final Four in 2008 and 2022. The biggest chances wasted were in 2010 and 2011. KU’s toughest opponent to make it to the F4 (and in fact the NC game) in 2011 was actually its R32 opponent Illinois. For 2010, KU would have been favored by 8.2 and 9.3 points for its would-be S16 and E8 games.

The 2012 path was Self’s fifth-best chance, which he converted by getting to New Orleans. 2018, which was a year in which KU didn’t get any seeding breaks or upsets, was the most-unlikely F4 for Self. Next we look at National Championship chances.

This table will look similar to the Final Four table. Note that the conditional formatting is set at Red for a 1/50 chance, Yellow for a 1/20 chance, and Green for a 1/5 chance. KU’s 2008 and 2022 titles are highlighted.

Self has had four really good chances, given the strength of his team and the draw he received (2008, 2010, 2011, 2022). But “really good” regarding odds of a title in the NCAA Tournament is still far less than even a 30% chance. He cashed in on two of those chances when just one title would have been above expectation. Aside from those four seasons, KU’s best title chance was below 10%. It’s really hard to win 6 games in the Big Dance, especially if you don’t get some breaks.

Removing KU’s team strength from the equation

Above there were 2 variables, the path that played out in that specific bracket and the relative strength of each KU team. As the 2008 squad was KU’s strongest, it isn’t unsurprising that it had better odds than other KU teams. What we will now do is remove the KU-specific component and look at each tournament path assuming a team with a 30.00 KenPom AdjEM is facing it. 30.00 is a very good AdjEM, and most teams that achieve that mark will earn a 1-seed. As good as KU has been in the past 20 seasons, its average AdjEM is 25.82, well below this 30.00 mark. It has only reached an AdjEM of 30.00 in 3 seasons under Self; 2008, 2010 and 2020.

This exercise will serve a dual-purpose. It is this 2020 team that didn’t get a chance to make a run for a title. With an AdjEM of 30.23, it is very close to the 30.00 number. We can plug the 2020 team for other KU teams in other brackets to analyze how successful it would have expected to be. For the first table, we see which paths were the easiest and which were the hardest to get to a Final 4. Later we will look at the National Championship table.

The 2022 path was the easiest, as that KU team faced weak S16 and E8 opponents. Unsurprisingly, 2011 was also a good break. However, none of these paths can remotely be described as a “cake-walk.” Even the 2011 path is expected to miss a F4 nearly half the time. On average, a KenPom team at 30.00 AdjEM (abbreviated from now on as KP30) is only expected to make 33.3% of Final Fours. Looking at the 10 years KU got a 1-seed, this KP30 team makes the F4 37.9% of the time. In the other 9 seasons (either as a 2,3, or 4 seed), this KP30 team makes the F4 only 28.1% of the time. Even the Final Four is flat-out hard to get to.

With National Championship chances, what first stands out is how much 2008 drops once you account for KU’s strength that season. The Final Four was loaded, with all four one-seeds. Having to beat North Carolina and Memphis was not easy. All told, a KP30 team getting KU’s path in 2008 was expected to win only 11.2% of the time. Incidentally, 11.2% was also the average chance of winning a national championship for a KP30 team.

KU’s easiest paths occurred in 2022 and 2011, with KU cashing in 2022. 2014, with teams like 11-seed Dayton, 7-seed UConn, and 8-seed Kentucky in the path; was an underrated good chance at a trophy. KU’s toughest paths have come relatively recently in the Self-tenure. The 2015 path would have been the only time KU had to face a chalk-bracket, as it would have had to have beaten three #1 seeds (including number-1 overall Kentucky) to earn a trophy.

The run of 2016-2018 also feels better in retrospect. While KU got three 1-seeds in this span, they only reached one Final Four and never played for a title in a championship game. But the cumulative expected titles a KP30 team would achieve in total for these three years is only 0.252. They turned out to be tough draws.

To tie this to the 2020 team, that group would have been expected to reach a Final Four, at best, 40% of the time. A National Championship would be 15% at best. Even as a 1-seed with a typical path, KP30 teams are likely to still get beat somewhere along the way. Winning 6 is tough.

Last, to highlight how difficult the tournament is, we’ll look at what AdjEM would be needed to reach certain thresholds. For instance, it would take an AdjEM of 34.72 for a team to reach the Final 4 50% of the time given the draws KU got over the past 19 tournaments. Even if we limit the draws to 1-seed paths only, the AdjEM would have to be 33.36. This is a level of play that only one KU team has reached in the 20-year Self-era. For National Championships, this level is more remarkable. To win a title 20% of the time, you’d need a 32.37 AdjEM. To win it 25% of the time, 34.00. To win it 33.3% of the time, the level of play would have to reach 36.30. To have even odds of winning, the estimated AdjEM would have to be 40.53. This type of performance has only achieved by the 1999 Duke team which was National Runners-up.

Off-Season Filler: Returning Minutes by Season

For 2024, KU’s roster will likely lose the following four players: Jalen Wilson, Kevin McCullar, Cam Martin, Gradey Dick. Of these four, three were starters in 2023 and logged serious minutes. Gradey is the most likely name to return despite having the highest draft potential, but as it sits today, we will project him going to the NBA after one season and leave him off the 2024 roster.

In the past 29 seasons, KU has returned 59% of its minutes played from the prior year. This is calculated by looking at a particular roster (say 2023), and calculating what percentage of possible minutes were logged by this roster in the season prior (which would be 2022 in the example). For 2023, KU returned 2022 starters in Jalen Wilson and Dejuan Harris. It also returned 2022 bench players such as K.J. Adams and Joseph Yesufu among a few others. Adding up these players’ 2022 minutes over total 2022 minutes, we calculate the 2023 roster had 37% of its minutes returning.

There is a small correlation between minutes returning and level of success. The R^2 is 0.2324. Judging success as the average game-score a certain roster produces, the best 5 Jayhawk teams in the past 29 seasons have been above the average returning minutes: 2008 (85% returning), 1997 (91%), 2010 (89%), 2002 (68%), and 2003 (62%). Of all the really-good Kansas teams, only the 2020 team (43% returning minutes) was noticeably below average. But even this is misleading. That 2020 team featured Udoka Azubuike, a senior by class who was injured for most of 2019. If you took Doke’s 2018 minutes played (the season in which he was mostly healthy), the returning minutes goes above 50%.

If we look at the correlation between returning minutes and NCAA success, there is a lower correlation (but still some with an R^2 of 0.0512). Looking at only the six Final 4 teams from this period, the teams had returning minutes of: 2022 (71%), 2018 (47%), 2012 (32%), 2008 (85%), 2003 (68%), 2002 (62%). In average, the Final 4 teams have 61% of returning minutes, slightly higher than non-F4 teams (58%). Again, low correlation here.

The 2024 outlook, as it sits today (3/22/23), would be that KU returns 52% of its minutes. If we project out this number to team success using historic rates over the past three decades, this would produce an average game-score of +9.14 (about a low-1/high-2 seed most years) and an estimated number of NCAA Tournament wins at 2.2. The following table has the data discussed in visual form. For the Factor column, this is the number of possible minutes played by returnees, with the highest possible number being 5.000. This column is divided by 5 to get the % column which gets color-formatted.

It should be mentioned that returning minutes aren’t the only thing used to project team success when CtH publishes preseason expectation numbers in October. Not all returning players are great, and sometimes newcomers add higher value than expected. With the transfer portal changes, a large share of a roster’s newcomers have proven themselves to be valuable college players, something that doesn’t get captured by this exercise. KU having Kevin McCullar in ’23 and Remy Martin in ’22 was valuable.

Kansas 71, Arkansas 72

Jalen Wilson had his 18th team-MVP game of the season in his final college basketball game. He recorded half of KU’s game-MVP’s this season, the highest rate of any Jayhawk since Frank Mason’s 2017 season. His game-score for the Arkansas game was slightly-above his season average. K.J. Adams had a fine game himself despite foul trouble. It was his best outing since the home Baylor game in mid-February. Dejuan Harris was the other Jayhawk starter to produce positive value, although he wasn’t mistake-free (5 turnovers counting the 5 second inbounds and 10 second backcourt violations). Gradey Dick was guarded well by Arkansas, and he missed some chances to have a bigger impact. His defense wasn’t terrible, but he needed to score more than 7 points. Kevin McCullar had some big baskets in the second-half, but his negative-value game was due to his defense. He was assigned to stop Arkansas’s scorers and had a rough second-half on the defensive side of things.

The bench finished the season with the distinction of being Self’s worst in his 20-year tenure. Joe Yesufu hit a ridiculous first-half 3 to finish slightly above zero, and Bobby Pettiford’s steal and layup also catapulted him above bubble-level. The trio of Udeh/Clemence/Ejiofor was not at all productive and may have contributed to the loss as they tried to fill in for K.J. Adams’ foul trouble. There were a lot of factors that went into deciding the outcome.

A questionable foul call on McCullar with 23 ticks left sent Arkansas to the line, where they would take the lead for good. Had the blocking call been a charge, KU would have likely gone to OT at worst.

Kansas 96, Howard 68

Kevin McCullar was KU’s most-valuable player in their Round of 64 win against Howard. The bulk of his value came from stingy defense and solid rebounding, but he also produced a positive offensive outing as well. Jalen Wilson had a quiet 20 points and also played strong defense, whereas Gradey Dick had the Jayhawks’ best offensive performance to overshadow his poor defense. This was also the bench’s first positive-value performance in its last 7 games.

The TEAM score of +15.94 was its third +15 game in the last four.

Kevin McCullar and the Value of a Replacement Player

KU plays Howard in the Round of 64 on Thursday afternoon, which according to the computer rankings turns out to be as good as a draw as they could have expected (Northern Kentucky is much better per KenPom, etc.). Even better news is the note that Kevin McCullar appears to be improving per Matt Tait. McCullar was not expected to play much in the Thursday quarterfinal game against West Virginia, but he actually went 26 minutes off the bench while fighting back spasms. This was perhaps a false sign of health, as he only played 12 minutes against Iowa St. in the Big 12 semifinals before reaggravating his back early in the second half. He would sit the remainder of that game and the entirety of the championship game.

If McCullar is healthy enough to go the rest of the way, KU has a much better chance to advance in any given round. Replacing a starter who has played in 71.6% of potential minutes this season would be a challenge in any season, but this is particularly considering the fact KU’s bench—particularly at the guard/wing position—has been historically bad this season.

Let’s start with Kevin’s value-metrics this season. Through 34 games, Kevin is producing an adjusted PPG +/- of +1.34, indicating that he is worth a little over a point per game vs. that of a bubble-level (very good) player¹. A comparable player to McCullar’s wing position who had a season around 0.00 (bubble-level) was Christian Braun as a freshman in 2020. So, if frosh Braun were replacing an injured McCullar this season, KU would expect to be about -1.34 per game worse.

But this isn’t what is actually happening, of course. The concept of a replacement player is employed as a way to judge player performance against a certain baseline. For a program as strong as KU, we expect any rotation player to be able to at least break-even when competing against a team on the NCAA Tournament bubble. After all, if you aren’t good enough to help Kansas beat a team on the bubble, you aren’t helping the program get to even a baseline level of what it deems success.

For 2023, KU’s backup guards and wings are (in terms of minutes): Joe Yesufu, Bobby Pettiford, and M.J. Rice. While they do play different positions, their roles overlap to a large degree. With McCullar out, any one of these three can be plugged into the rotation without much change in style (unlike say, any of the reserve bigs). This matters, because when KU goes into its bench on the perimeter, it gets quite a bit worse. Plugging in Udeh for Adams has been less of a drop off, and in some regards (size, rim protection) it has been an improvement. In the season-summary page (link here), a table listing player value in 3 different ways is shown. As of 3/13/2023, this is what it looked like for Kevin McCullar.

This shows that Kevin is producing +1.34 more net points per game for Kansas than a hypothetical bubble-level player at his position (again, think 2020 Braun). The next column translates the per game value to a Per 100 possession level. At +2.51, this number is still quite solid. The final column looks at Wins Above Replacement, not over that of a bubble-level player but of an even worse replacement-level performer. WAR is calculated against a baseline of an average D-1 player. Clearing the value provided by an average D-1 player is something that should be very achievable for any KU player to make the top 8-9 in the rotation. At +2.62 WAR, McCullar has played well enough this season to add this many wins to his team over that of the average D-1 basketball player.

Now we will show the same table as before, just with Pettiford, Yesufu, and Rice added as a comparison.

Kevin’s minutes weren’t replaced by a 0.00, bubble-level player at the Big 12 Tournament. They were replaced by these three. In the final two games of the Big 12 Tourney, these three combined for 93 minutes. At 46.5 minutes per game, this was considerably more than the 28.2 that these three average. And this was with McCullar playing some during the Iowa State game. For the Texas game alone, the bench guards/wings played 56 minutes as they covered McCullar’s missing 28 minutes. In these two games, collectively this trio produced a mind-boggling -17.42 points of value. While KU won against Iowa State, it was in spite of the bench. With the Texas game, the bench’s performance only heightened the poor outing of the normal starters.

If we take the difference between McCullar and one of his replacements, we see that Kevin is a better option at the guard/wing by anywhere from 2.21 to 3.21 points per game. The Per 100 numbers are even more stark. Over the course of 100 possessions, Kevin is +6.42 when compared to Joe and +11.00 when compared to Bobby (M.J. Rice is about equal to Bobby). This difference might be able to be masked over a few possessions, but over the course of a game, and certainly over the course of a tournament run; this drop in value will rear its ugly head.

If Kevin were to be out for the rest of the year, KU would need to replace about 50 possessions per game. Even if you could use, perhaps 60% of those minutes with Joe (who has been the best of these three), you would still expect to lose about 4.1 points of value per game. This would dramatically shrink KU’s chances at making a deep run barring some miraculous play from one of the bench reserves. KU needs Kevin, and in fact all of its starters, healthy. This year more than others.

Since 1994, the first year any player-value scores have been estimated, the average SG/SF (wing) player-value for KU players who played at least 40% of available minutes is around +1.46. There is a wide variety of names on this list: Haase, Gregory, Langford, Rush, Wiggins, Mykhailiuk, Grimes, and Agbaji. Some good, some great, some not. McCullar is right around average, which given the comparison still makes him an excellent college basketball player.

But his value to the 2023 team is amplified by who replaces his minutes when he cannot play.

1- As explained, a bubble-level player would be the average performer for a bubble team. A bubble team is determined by looking at the median-level adjusted efficiency at-large team. There is a difference between a team’s adjusted efficiency and its resume, but there is also large overlap. A team around +14.00 on KenPom is normally close to the bubble.

Kansas 56, Texas 76

Only two Jayhawks who played were above-bubble, Jalen Wilson and Joe Yesufu. Jalen had an uncharacteristically poor defensive game, otherwise his value-score would have been much better. Joe played 29 minutes filling in for injured starter Kevin McCullar, and finished in double-figures with 11 points.

The remaining starters had poor outings, but none were terrible. Gradey Dick shot poorly but defended adequately. Dejuan Harris likewise had a bad offensive game, but his defense was solid. K.J. Adams wasn’t great, but he outplayed freshmen Ernest Udeh and M.J. Rice. The bench has now had 6 consecutive below-zero outings for Kansas, dropping it to the worst bench in the 20-year Self era.

Next week begins the NCAA Tournament. One hopes the team can use this poor outing as motivation to come out focused and win two in Des Moines next week. It was the TEAM’s second-worst outing of the season (and 8th below-zero game overall).

Kansas 71, Iowa State 58

Jalen Wilson had his second-best performance of the season, and Dejuan Harris continued his torrid play of late (averaging +5.28 over his last 10 games) to lead Kansas into the Big 12 Tournament finals. K.J. Adams had trouble inside, giving up a few put-backs to hurt his value score. KU won despite having to go deeper into its bench with Kevin McCullar playing limited minutes due to injury. The bench guards/wings were -10.39 collectively.

Seed Performance in the NCAA Tournament

Since 1985, the NCAA Tournament has had 64+ teams, with four regionals of teams seeded 1 – 16. In this format, each team has to win 6 games to win the Tournament with each seed having a different path (for instance, a 1 seed has to beat a 16, 8/9, 4/5/12/13, 2/3/6/7/10/11/14/15 to make the Final 4…this differs from a 2 seed’s path). What this means, is that given this many instances, we have some interesting data by seed-line.

For today, we will focus on the prevalence of each seed in the Final 4. With four regionals, the Final 4 is made up of the four regional winners. It isn’t until this round that teams of the same seed can meet. We will first look at which percent of each seed makes the Final 4.

Above is the table, and below is the bar chart of the same data.

The 1-seed is the most prevalent (unsurprising), with a declining frequency by seed down to the 7 seed. However, we see a jump when we get to the 8-seed, and the 11-seed has also won its regional more times (5x) than the 6-seed (3x) since 1985. This is definitely a surprise in the data. Some might hypothesize that the reason for this is that the 11-seeds often contain undervalued mid-majors, whereas the 6-seeds are often weaker power conference teams, and that the best mid-majors are therefore under-seeded. In other words, it would be a claim that the bracket-makers get things wrong along these middle seed-lines.

However, this doesn’t really work. After all, the 6 v. 11 game is played 4 times in the First Round. If 11-seeds were systematically under-seeded, it would show up here. Instead, in 148 games, we see that 6-seeds win 61% of these matchups and have an average scoring margin of +2.96 (nestled in between the 5v12 and 7v10 margins). For whatever reason, the 11 seeds have had more success than the 6’s at winning a regional. However, a 6-seed has won the NCAA Tournament (1988 Kansas). Of course the 1983 N.C. State championship team was also a 6-seed. It falls outside the range of 1985-present, but their run was no less impressive than it would be today.

But now let’s look a little deeper at the 1-seed’s frequency at a Final 4. In the 37 tournaments since 1985, 41% (60/148) of the top seeds made the Final 4. This works out to an expected number of 1-seeds per in the Final 4 per tournament of 1.62. Because each regional is decided independently, this is just the average expectation. Some years it might be 1, some it might be 2. And other years, less frequently, it might be 3 or all 4. Or, it could be 0.

Historically, there has been 1 tournament (from 1985-2022) where all four 1-seeds made it to the Final 4. This was in 2008. There have also been 2 tournaments in this span where no 1-seed has made it to the Final 4. This happened fairly close to 2008, with 2006 and 2011 seeing no 1-seeds win their regionals.

We are left with 148 independent events, which are the historic regional results by seed. Using this data set, we have had all four 1-seeds win their regionals in exactly 1 tournament of 37 (2.7%). How often should we expect this to happen? If we take (0.4054)^4, we get the answer, which is also 2.7%. While one would expect the 1-seed to have the best chance of winning its regional, having all four 1-seeds go 16-0 before the Final 4 is a rare event, as shown by the historic results.

The flip side is the question of how frequently we should see NO 1-seeds make the Final 4. It has happened twice in 37 tournaments, but is that about what we’d expect? The formula here will calculate how likely it is that at least one 1-seed makes the Final 4. It is similar to the earlier example, and it is calculated thusly: 1 – ((1-0.4054)^4). The result is 87.5%, meaning that we’d expect to see no 1-seeds about 12.5% of the time (roughly 1 year out of 8). In reality, this has only occurred 2 years out of 37.

Using the 1-seed success rate at making a Final 4, the last question is, what will happen next? All 1-seeds, or no 1-seeds? Clearly, the answer is the latter is more likely. This does depend on the strength of the 1-seeds relative to the field. In 2008, the 1-seeds were all excellent (although 2011 had a good crop of 1-seeds by computer metrics as well). But assuming the relative 1-seed skill-level will remain what it has been over the past 37 seasons, the math says it is 5.15 more likely that all the 1-seeds miss the Final 4 in any given year than it is that all four 1-seeds make the Final 4.

We will next look at other seed-lines. While 2-seeds are all excellent teams, their performance when compared to the 1-seeds is a significant drop-off. We’d expect all four 2-seeds to make a Final 4 only once in 458 tournaments. For the 3-seeds, this would occur once in 5,744 tournaments. This gets exponentially rarer as we move down the seed-line.

Last, let’s look at how often we’d expect to not see any top seed. This table has the estimates:

2011 is the only tourney to fail to have a 1 or 2 seed in the Final 4. The other two scenarios haven’t occurred and likely won’t in your lifetime.

Kansas 59, Texas 75

In a game that meant nothing to the conference race, which KU had already clinched, the Jayhawks played flat in Austin and got beat on Texas’ senior day by 16. Jalen Wilson had another great outing in a game when his teammates played poorly. The TEAM’s Game Score of -6.19 (meaning this performance was 6.19 points worse than the expected performance for a bubble team) was its third-worst game of the season and its 7th negative game out of 31.

In these 7 games, Jalen Wilson is averaging a game score of +9.00, whereas the others are collectively at -15.94 for the season. Jalen is having a good season, his current Adj. PPG +/- number is +5.53, but the fact he has raised his game to a high level when others aren’t picking up the slack is impressive. He helped KU win a few of these poorly played games earlier in the season (Southern Utah, Wisconsin).

Gradey Dick had his worst game of the year and has produced negative value over his past 10 games. On the flip side, Dejuan Harris is showing he can provide value even in games where he shoots poorly. Despite managing just 5 points and multiple missed shots, Harris assisted nicely (without turning it over) and defended strongly against Texas’ veteran guards. Kevin McCullar also had a nice defensive game to end slightly above-bubble. K.J. Adams was slightly below-bubble.

The bench had another poor outing, as Bobby Pettiford played the worst of the lot. Only Naadir Tharpe (2012), Justin Wesley (2012), and Stephen Vinson (2006) have had worse Per100 seasons than Bobby out of all rotation players since 2006. (M.J. Rice is also on this list, right next to Pettiford). But Bobby’s played the most out of these names, at 27% of available minutes.