Sports and Numbers

Current Projected Postseason

NCAA Tournament: #1 Seed (1st overall) - Actual seed

(As of 03/17/10.  Source: Bracket Matrix used throughout season)

Player Contributions - 2009-10 Season
Latest Game Data Included - vs Kansas State (Big XII Tournament)

Total Impact (ePSAN) - This represents the players with the greatest total impact on the team thus far.  It adjusts for quality of the opponent, and it gives more weight to recent performances.  It is broken down into offensive (ePSAN-O), defensive (ePSAN-D) and composite or overall (ePSAN-Comp).

Efficiency (ePSAN70) - Takes the "Total Impact" ratings and expresses them as "per-70-possessions," which is the number of possessions in a typical college basketball game.  Essentially adjusts for playing time.

Key Player Analysis
Latest Game Data Included - vs Kansas State (Big XII Tournament)

One way to describe the key player on a team is to say that he is the player who contributes most positively, whose contributions make a difference and whose performance does not tend to get worse against better opponents.  That's what I'll attempt to ascertain with this ongoing analysis.

Note: This used to be called the "Most Valuable Player" analysis, but I have since come to feel that is misleading.  The most valuable player would be the one who has the highest impact rating, in theory.  This analysis finds a player who may or may not have the highest impact but whose impact appears to make a big difference in the team's performance.  Even a star player's off night can be compensated.  But the "key player" in this analysis will often be the guy whose bad night cannot be compensated for well.

Methodology: Using multiple regression analysis, calculate which players' impact ratings for each game (PSAN-Comp) best explain the variability in KU's team performance (using power ratings to adjust for opponent strength).  That is, which player(s) ups and downs best explain KU's ups and downs.  Remove players who tend to perform worse against better competition.  Among the remaining group, the player with the highest "Total Impact" rating for the season (see above, ePSAN-O/D/Comp) is designated "Key Player" for respective category.

1) Multiple regression currently shows these players' ratings to be significant in explaining KU's team performance variability:

2) Eliminate those players whose PSAN-Comp game ratings have a correlation in the wrong direction with strength of opposition.  That is, a good key player should not play worse against better competition.

Stat geeks: I will use the correlation coefficient (r) and a significance test at the p=0.20 level.  This means eliminating a player whose correlation coefficient would be so significantly pointed in the wrong direction that there is only a 20% probability that it is bad due to chance.  Why do I use this level instead of the typical scientific 0.05 level?  It's basketball, not curing cancer ... and I like the "80/20" rule, as they say.  I calculate the p-value by using this formula  and using the one-tailed test, because I only care about whether the coefficient is pointing in the wrong direction.  The wrong direction for offense and defense is actually a positive correlation.  Why?  Example:  A lower offense against a lower defensive rating means a player does worse against stronger defenses, since for defense, negative numbers are better. Same is true for offense, in reverse.  For composite, however, a negative correlation is the wrong direction.

Remove players whose "r" is in wrong direction (+ for offense/defense, - for composite) and has a "p" value corresponding to the shown "r" that is lower than 0.20.  Note: This is equivalent to removing any players whose defense/offense/composite appears as a "Disappear" in the "Step Up Index" below.

3) Highest ePSAN-O, lowest ePSAN-D and highest ePSAN-Comp in list from #2 above is given designation of Key Player for category.

Team Performance Trends

Latest Game Data Included - vs Kansas State (Big XII Tournament)

This graph will show how the team is performing on offense, defense and overall as the season progresses.  The "Zero" value represents the NCAA average, so +40 for offense means 40 pts/100 possessions better than the average, while a -25 would mean 25 pts/100 possessions better on defense, since it is better to allow fewer points on defense.  The dotted lines represent the best-fitting line, representing either improvement or worsening.  Overall performance adjusts for the fact that negative is better for defense, so overall is better when higher.  NOTE: Discrepancies between this chart and the ones you see in game previews arise because the previews are based on KenPom's game plan numbers, which provide slightly different numbers for tempo and efficiency (not at all significant).

ANALYSIS - After five straight games with overall ratings below the trendline, KU has now played better than the trendline in four of the last five, with the Texas Tech game in the tournament quarters being the lone exception.  The defense has been better than its trendline in the last four games.  Still, what's troublesome is that the season-long trend for both defense and overall is statistically signficantly downward.  Maybe it's not a terribly steep slope, but it's almost certainly there.  The offense looks like it is slightly upward, but it doesn't meet the criteria for statistical significance.  Thank goodness the team has put it together a bit in the last five games, otherwise it was clearly the picture of a worsening team before that.