KU falls to 31/48 in DPPI

After the drubbing by Texas this weekend, KU has fallen to 31st in the DPPI Power and 48th in the DPPI Performance ratings. Here is a brief explanation of these two ratings.

DPPI Power -- The rating based on the expectation of victory in a head to head matchup. KU is 31st, meaning they would be favored against everyone lower than 31 and the underdog against everyone higher than 31.

DPPI Performance -- This is derived by doing a round robin pitting all 357 teams against each other and summing up the expected victories from those 356 games. For instance, KU's DPPI Performance rating is 0.8343 meaning they would win 83.43% (or about 297) of those 356 round-robin matchups.

So you may ask, why would the not win 326 and lose 30 which is the number of games they would be favored by and underdogs in? That is because each team, in addition to the Power Ranking which essentially sets the projected scoring margin, has a distribution function based on their past performance. Okay, that is a hyper technical word. Let me rephrase. Basically, I take the past performance in the games already played and measure how consistently the team played to expectation and how much they played considerably above or below expectation. In other words, how consistent is a team to playing close to the expectation. In addition, when simultaneously solving this distribution function for all 357 teams, there will be non-zero means in the distribution function. Okay, super nerdy word that so let me try to simplify. The gist of this is that all teams are inconsistent to some degree. If you average out over all 357 teams, the average says that the typical team will play within 10 points of expectation 95% of the time. Any time a team plays more than 10 points above or below expectation, it represents some sort of anomaly, if you will, that needs to be dealt with differently. Also, if you average all 357 teams for all games, both the times the play above expectation and the times they play below expectation, the average for all games will come to zero. There will be exactly as many points scored above expectation and below expectation. But that is not necessarily true for any given team. When you are simultaneously solving for 357 teams, some teams will average out above zero and some teams will average out below zero (non-zero means). Treat this sort of like a long-term trend for the team.

So every team has one of these distribution functions. This distribution function is the input to calculating the probability of victory. The two variables in this distribution function are variance (or standard deviation) and mean. Once you have this distribution for each team and you have a projected scoring margin (from the DPPI Power Rating), you can calculate the probability of a team winning.

So basically, here is the gist of the why consistency matters. A team that is very inconsistent (KU is one of the least consistent teams in the country right now) will have a high variance in their distribution function. What this means (and here is the central most important thing to understand) is that a team with high variance will have a lower probability of winning with a given projected scoring margin. So for instance, hypothetically KU may have a projected scoring margin against team X of 3 points and Houston may have a projected scoring margin against team x of 3 points as well. But KU's probability of winning is 60% whereas Houston's probability of winning is 80%. KU has a lower probability of winning because they have a high variance (they are more inconsistent). So when you do the round robin against all 356 of the other teams, KU will have a lower probability of winning each game than most other teams. And the total number of expected wins is the sum of the 356 individual game probabilities.

So the power ranking is really the best ranking for looking at how KU compares to other teams in any given game, but the Performance ranking starts to be better over a season. Either way, KU is not playing like a top 25 team in Power or Performance. That is alarming. We now have pretty close to a zero short-term trend, but we still have a pretty decent long-term trend. The long-term trend is included in the calculation of the Performance rating. This year is looking quite tenuous for KU unless they start to get better play in the paint.

Okay, so maybe a picture will help. The picture below is the probability distribution functions for KU and for all of Div. 1. You can see that KU's probability of winning is less than the average Div 1 team for any given scoring margin. Hope this is helpful

Probability Distribution Function