Micro-Analysis #2: KU

While we're waiting for the next installment of the Quality Curve Analysis, I thought another micro-analysis article would be fun. I had a couple of teams from which to choose, and from this particular team, I had three different ideas on how to approach them. I ultimately settled upon what you are about to read, so let's dissect KU in the same manner they've dissected a lot of their opponents this year.

Can Kansas Continue Consistent Outcomes Conducive to a Championship?

For a number of reasons, I believe the answer to this question will eventually be no come tournament time. First off, historical analysis shows that only one defending champion since 2002 has advanced beyond the S16 (2007 FLA), and they returned approximately 95% of their previous team's production. In 1998, ARI advanced to the E8, and from their 1997 National Championship team, they returned close to 90% (I personally cannot validate this figure due to lack of reliable data for that time period, but I do know it is close to that value). Apparently, returning 90% of your team is the secret to overcoming this barrier. However, 2023 KU returned approximately 30% of their 2022 National Championship team. Second off, power ratings (elo-based ratings systems) are starting to lower their probabilities for a deep tournament run. Elo ratings systems are based on margin of victory, so wider margin of victories imply higher probabilities of advancement. As we will see below, KU has squeaked by a lot of average teams, so their low margin of victory has begun to punish their elo rating. Instead of relying on these macro-models, let's use micro-analysis to answer two questions: How does Kansas win and can they keep it up?

First, I want to do a simplistic win-correlation model (not to be confused with advanced metrics like win share percentage ratings). Then, I want to break down the important factors further on a player-by-player basis. Even though KU has played 21 games this season, I am only analyzing power-conference opponents (16 total, now-power conference teams distort the figures, and let's be honest, KU should beat these teams easily).

This is a chart of four-factor analysis, and as usual, I slam a lot of information into one chart. Blue is KU data, Orange is neutral/shared metrics, and Black is opponent's data. MOV is Margin of Victory, PPP is points per possession, P# is number of Possessions, EFG% is effective field goal percentage, FTD is Free Throw Differential (KU FTs minus Opps FTs, positive being good, negative being bad), OR% is Offensive Rebounding Percentage, TOD is Turnover Differential (Opp TO - KU TO, positive being good for KU, negative being bad for KU), and TOR is Turnover Rate (TOD / P#). All correlation values (CORREL row) are being compared to margin of victory because we want to see how KU wins and/or loses.

Here are my thoughts:

1. Overall, KU's offense is more important to their wins than their defense. KU Points show strong 0.8494 correlation factor compared to the -0.5144 of Opponent Points (Keep in mind, the closer to +1.0000 and -1.0000 matters for CORREL, the closer to 0.0000 means no discernible relationship between the data and MOV). Even when pace is standardized with PPP, KU's PPP is more correlated than the opponent's PPP. I've said in a previous article that KU will struggle to defend the paint this year. As this data seems to show, KU is going to give up a predictable level of points, but the difference in winning and losing will come to down to KU's ability to outscore.
2. When we look at the breakdown of the Four Factors (like we did in the CONN article), we see a common thread: Offense is still more correlated to MOV than defense. EFG% has a offensive CORREL of 0.7602 compared to a defensive CORREL of -0.5115. Those are both fairly strong correlations, but the offensive metric is closer to the +/- 1.0000 value, so its significance to wins and losses is stronger. The same goes with turnovers, -0.4200 on offense versus a 0.2709 on defense (higher turnovers means lower MOV, which is why this CORREL is negatively related).
3. Offensive rebounding percentage is even less correlated to margin of victory, but slightly more important on the defensive side of the ball. I cannot logically explain the offensive side, but there is one rational explanation for the defensive side. If you are a team that has to outscore your opponent to win the game, second chance points make this challenge even steeper. In the six games where KU gave up an Opp OR% higher than 0.296, four of those six games were losses, and the other two games were a 2-pt and 5-win, respectively. In their most recent game at UK, they allowed an astronomically low 0.074 Opp OR%, and if you watched that game, it definitely had an impact on the outcome considering UK's struggles this year in scoring the ball.
   
4. I'm not even sure what to make of free throws. First, you know I don't like the Free Throw Rate component of the Four Factors. Second, free throw percentage on a game-by-game basis can be easily skewed by a lower number of attempts (i.e. 2/4 = .500 whereas 3/4 = .750, and if you're only taking four free attempts, free throws probably don't matter to the game's outcome). On average, KU's opponents make two more FTs than KU, and the 5.6 standard deviation makes me nervous as well, but this probably has something to do with the +12 outlier in the OU game.

Now that we know that offense is more important than defense to KU, let's see how individual player production matters to KU wins and losses with another correlation chart.

Each value represents the correlation coefficient of that player's game stat to the game's margin of victory. Let's do a player-by-player analysis, followed by the points stat as a standalone subject.

• Jalen Wilson: His two most influential categories are Minutes Played and Defensive Rebounding. If KU is winning big, his minutes go down, and if KU is losing or tied, he gets every playable minute, which is why this correlation is negative. If he is getting a lot of defensive rebounds, then the opponent is missing shots, and every missed opponent shot is a blessing for KU. The majority of his correlations are inverted from the rest of his teammates (his are positive when theirs are negative, and vice-versa). It is correlation's way of saying that "he is going to his numbers no matter what happens". Unless a team has a true lock-down defender at the 4-spot, I don't think game-planning against Jalen is a good idea, and the rest of the numbers seem to confirm.
• Gradey Dick: There's no stat where Gradey is most significant, either positively or negatively, in correlation analysis, but he's usually 2nd or 3rd in every category. I find this result strange because he's essentially the team's second best scorer, so if you can't stop Jalen, logically he seems like the next best target. I did lightly emphasize the significance of his 3P%. While it's slightly behind the importance of McCullar's 3P%, I did examine Gradey's shooting splits for a potential micro-analysis article, and they point to a cold shooting night from Gradey spelling doom for KU, but the sample sizes were too small for the probabilities to be reliable.
• KJ Adams: His most significant contributions are offensive rebounds (by a long-shot) and points (roughly a three-way tie). Offensive rebounds are extra chances at points for KU, which is exactly what a must-outscore-the-opponent team like KU needs. Digging deeper into points, his 2PM and 2P% are second-important to the team. For a player whose interior-offensive presence has a roughly 0.5000 correlation to wins and losses, it is unfortunate he is also tasked with the responsibility of guarding the opponent's best post player. He has played a lot of games (or missed a lot of minutes depending on how you want to word it) in foul trouble, so his lack of presence on the floor will likely be one of the factors in a KU tournament loss.
• Kevin McCullar: Now we get to one of the two players dominating the correlations, and Kevin's FGA, 2PA, three-point shooting, and steals are the most significant to the team. Remember, correlation doesn't mean causation. This analysis doesn't mean Kevin needs more shots in the offense to guarantee a win, but instead, when he is not getting his average shots per game, KU is either losing or in a close one. As for the explanation, higher quantities of shots could imply that he is active in the offense or the opposing defense is focused on limiting shots from other players. On the contrary, shot attempts in all three categories (FGA, 2PA and 3PA) show the smallest correlation value, and this makes sense because shot attempts don't increase your team's score.
• DaJaun Harris: Offense starts with the point guard, so if a team can take down Harris, they will have a good chance at taking out the whole KU offense. His production in FGM, FG%, 2PM, 2P%, FT%, and TOs are the most correlated to wins and losses than any other KU player in these categories. Furthermore, his correlations in these categories are the highest among all correlations. In the other stat categories, the highest correlations range in strength from 0.3730 to 0.5140 (absolute values), but Harris' correlation in five of his six categories are in the 0.6000 range, indicating the strength of this connection. He also has the second-highest correlation to points, which is strange since he is more of a distribute-and-defend point guard. How can an opponent neutralize Harris? TENN kept Harris in foul trouble, producing season lows in minutes and assists. IAST's and TCU's team-defense forced the ball out of Harris's hands, producing season-lows in FGM, FGA, FG%, 2PM, 2PA, 2P%, and points. The guards from WISC, KNST, and BAY kept active pressure on Harris, resulting in season-highs for TOs (as did TENN).
• Points: I was actually surprised to find that the point totals of three players (rather than one or two) would correlate to margin of victory. I also have a plausible theory to its meaning: KU is missing a third-option on offense. If you count Wilson and Gradey as options #1 and #2, respectively, then who is option #3? Adams gets a lot of his production via offensive action (assisted shot attempts and offensive putbacks), so running a lot of constructed offense for a third option doesn't seem plausible, especially since KU already runs a significant portion of their offense for Wilson and Gradey. The next choice is McCullar, and correlation analysis shows KU does better when he gets more looks. However, his shooting percentages leave a lot to be desired, and the correlation analysis says only his 3P% correlates with wins and losses. This likely means he is best-suited as a weak-side shooter for spacing and/or countering post double-team tactics. This leaves us with Harris, and a PG would make sense as a scoring option for any offense (in my opinion, they should be one of the first two options). However, Harris is the fifth highest scorer, behind the aforementioned four, and of the three starters returning from last year's team, he has shown the least year-over-year improvement. IF KU is going to make another tournament run, or at least past the S16, 1) They need to score, 2) They need a third-option in scoring, and 3) They need the third-option to be Harris and the strength of his correlations to margin of victory compared to other players demonstrate why it has to be Harris.
  

KU, on the whole, are these five players, and very little else. They account for 99% of the starts, 77.4% of the minutes (which is an extreme number), 82.6% of all field goal attempts, and 84.1% of all points. Their scoring averages are, in order, 21.4ppg, 14.8ppg, 10.7ppg, 10.4ppg, and 7.2ppg. To further illustrate the magnitude of their starting five, the sixth man off the bench has a scoring average of 3.3ppg on 11.8 minutes. With almost everything in KU's offense accounted for by the Starting Five, where are the extra points coming from when needed? If the Bench players don't get their allotted minutes, then the starting five need to hit their averages and cover the Bench's deficit (15.9% of KU's total point average). If one or two of the starters have a bad night, can the rest of the starters plus the bench hit their averages while picking up the slack? Finally, every team needs to take their best to the next level for the tourney. Can KU take their seemingly already stretched numbers to a higher level for another tournament run?

If I didn't bog you down with a swarm of numbers, stats, and metrics, I hope these questions were answered by this second foray into micro-analysis. As always, thanks for reading my work.