Tuesday, January 22, 2008

Pythagoras at the 30.6% Mark

If the Big East regular season ended today Georgetown would take the regular season title and earn the #1 seed in the Big East Tournament. For anyone who read the preseason previews this should come as no surprise. The surprises are the #2 - #4 seeds -- the three other first round bye seeds. As of today (Monday) those teams would be (I don't know how tie breakers would apply in this situation, given the teams have played 6 games) Cincinnati, DePaul and West Virginia. A few prognosticators, who felt Bob Huggins would not have a long learning curve with the Mountaineer squad, might engage in some self-congratulations, but no one outside of the hard core fan base would have placed the Blue Demons and Bearcats in the top quartile.

I decided to see how things would settle if I applied the Pythagorean Winning Percentage calculation to the teams. Ken Pomeroy uses this formula as part of the ranking system he implemented several seasons ago. Each D1 team's winning percentage appears in the far right column of his Stats Page. I decided to use only the points for and against from Big East play, a practice Pomeroy used for the first 2 or so years that he used the stat. He eventually abandoned this in favor of the season-long winning percentage. The result are shown in the table below. Results are presented in the projected order of finish.



Actual RecordProjected Record
TeamGmsWLPctWLPct
Louisville5320.60013.14.90.729
West Virginia6420.66713.05.00.722
Georgetown5410.80012.95.10.715
Pittsburgh5320.60012.85.20.713
Providence College5320.60011.36.70.630
Connecticut6330.50011.16.90.618
Marquette6330.5009.98.10.550
Villanova5320.6009.88.20.547
De Paul6420.6679.78.30.540
Syracuse6330.5009.28.80.511
Notre Dame5320.6007.910.10.442
Seton Hall5230.4007.810.20.432
Cincinnati6420.6677.510.50.417
St. John's5140.2005.912.10.328
South Florida6150.1674.713.30.263
Rutgers6060.0001.216.80.069

Those who have been wondering what happened to their preseason favorites should find comfort here. Louisville, the projected regular season champ and #1 seed in the BET in this projection was considered a co-favorite with the Hoyas in most preseason polls. Early injuries and (continued) internal strife have, according to common wisdom, hurt their chances to finish among the bye seeds. The balance of the bye seeds here were unanimous (Georgetown, Pittsburgh) bye team selections or present on a few ballots (West Virginia). Other results which coincide with common perceptions so far include:
1. Rutgers and St. John's will struggle. Those two, along with less obvious candidate South Florida, are slated to finish among the bottom three.
2. Cincinnati will finish in the lower division. The projection shows the Bearcats, despite their current 4-2 record, will finish #13 with something like an 8-11 record. The high number of wins, relative to their projected place of finish, is evidence of the "win bias". The #13 team will most likely win 3 - 5 games, making the Bearcats (with 4 already) unlikely candidates for the bottom quartile. If Cincinnati does win 8 games they will finish somewhere between #9 and #11 most likely.

The win/loss breakdown (for those who counted) is about 148/140, overstating the number of wins. The breakdown should be 144/144, not 148/140. Indeed if rounded to the nearest integer, the process applied predicts that 9 teams will win 10 or more games, a very unlikely possibility (the bottom 4 - 5 teams would have very few wins among them). Bill James, originally applied this formula for predicting the order of finish for major league baseball teams during the season. Ken Pomeroy modified the exponent James used (as he explains in his 2004 blog entry "Meet My Friend Pythagoras") to adapt the formula to D1 basketball. Pomeroy compared actual winning percentages to Pythagorean winning percentages to identify lucky and unlucky teams.

No comments: