Tuesday, November 6, 2007

Gauss and Gasaway, Pomeroy and Pythagoras


Over at the Basketball Prospectus John Gasaway of The Big Ten Wonk fame and Ken Pomeroy of kenpom.com have worked together to produce previews of many of the D1 power and mid-major conferences. They are following the model used by Pomeroy when he previewed the Big 12 Conference going into the 2006-07 season. Their analysis relies on many of the possession-based stats that Dean Oliver developed for analysis of the NBA and which Pomeroy adapted for the D1 game. While the individual team analysis highlights (statistically) an area in which the team excelled, the "common denominator stats" both analysts employ include the Pythagorean winning percentage developed by Bill James and adapted to the D1 game by Pomeroy, the returning PT% (focuses on year-to-year continuity). I have waited patiently for their Big East preview, but when Pomeroy published the A10 preview last week, I decided to pull those numbers together and take a look myself. Here are some of the things I found...

ActualPythagorean Proj.
TeamWLPct.WLPct.Diff
St. John's790.4384.811.20.2982.2
Marquette1060.6259.07.00.5621.0
DePaul970.5638.47.60.5220.6
Syracuse1060.6259.66.40.6010.4
Notre Dame1150.68810.75.30.6690.3
Pittsburgh1240.75011.84.20.7350.2
West Virginia970.5639.07.00.5650.0
Georgetown1330.81313.12.90.817-0.1
Villanova970.5639.16.90.571-0.1
Rutgers3130.1883.112.90.191-0.1
Louisville1240.75012.23.80.764-0.2
South Florida3130.1883.312.70.207-0.3
Providence880.5008.77.30.542-0.7
Seton Hall4120.2505.110.90.316-1.1
Connecticut6100.3757.28.80.447-1.2
Cincinnati2140.1253.612.40.225-1.6

I used the standard formula developed by Bill James, as modified by Ken Pomeroy over the past several years. By using the "points for..." and "points against..." I departed from the modification that Pomeroy used last season to develop his new rating system. I used 9.6 as the exponent to derive the winning percentage. Note the "unluckiest" teams last season were Cincinnati and Connecticut, very reasonable candidates given the inexperience of the teams (Connecticut returned only 20% of their previous year's playing time) and in the case of Cincinnati, the Bearcats broke in a new coaching staff in addition to the squad. St. John's as the "luckiest" team is something of a surprise. I would have expected a veteran team like Syracuse or Pittsburgh to be the one to take full advantage. Projecting that winning percentage for an 18 game schedule like the one to be played this year yields these projected wins for each team:

18 Game
TeamPTFPTAPct.WL
Georgetown10809240.81714.73.3
Louisville11209910.76413.84.2
Pittsburgh10799700.73513.24.8
Notre Dame120311180.66912.06.0
Syracuse115011020.60110.87.2
Villanova109810660.57110.37.7
West Virginia111510850.56510.27.8
Marquette112911000.56210.17.9
Providence115311330.5429.88.2
DePaul9919820.5229.48.6
Connecticut102710500.4478.010.0
Seton Hall113412290.3165.712.3
St. John's98810800.2985.412.6
Cincinnati100411420.2254.113.9
South Florida98411320.2073.714.3
Rutgers93210830.1913.414.6

The first two columns are each team's "points for..." and "points against..." in conference play only. The 3rd column is the team's computed Pythagorean percentage based on the James/Pomeroy formula. And the last two columns project each team's 18 game record. This looks much like last year's final standings, though there are several exceptions. These rankings suggest that Pittsburgh and Notre Dame, for example, will finish 3th and 4th (respectively) again this season. Given the player losses, Gray & Kendall, Carter & Falls, how likely is it? Syracuse lost their front court (and Andy Rautins), Villanova graduated three seniors and West Virginia changed coaches (and playing styles). The table below juxtaposes the Pythagorean projected wins from the previous table with the percentage of returning minutes played.

18 GameRet.
TeamWinsMin%
Georgetown14.778.7
Louisville13.882.3
Pittsburgh13.259.1
Notre Dame12.067.8
Syracuse10.838.8
Villanova10.353.0
West Virginia10.271.0
Marquette10.194.0
Providence9.884.1
DePaul9.450.9
Connecticut8.098.9
Seton Hall5.776.9
St. John's5.443.6
Cincinnati4.174.9
South Florida3.752.7
Rutgers3.459.9

The returning minutes add a bit more to the picture. Keeping in mind that Connecticut suffered a net of -8 wins from 2006 to 2007 (14-2 to 6-10) when they returned 20% of their playing time. Syracuse had the benefit of being slightly lucky (and very consistent) last season -- will they be able to win 11 games this season? Even with arguable the best entering class that may be too much to expect. Notice that Coach Boeheim played a relatively short rotation in the St. Rose game, choosing to allocate more minutes to 6 - 7 players, rather than run 10+ players through the game. DePaul, like Syracuse brings a very good class to the court this season, while losing nearly ½ of their PT from last season. 9 - 10 wins, despite the fact that they lost their two main scorers, may not be unrealistic. Connecticut returns virtually all of their playing time from last season, one in which they were both very inconsistent (they were after all under classmen) and very unlucky. Recouping all 8 (or if you consider proportions, 9 wins...) wins may be over optimistic, but the Huskies will certainly do 3 - 4 wins better than last season. And that would put them into the NIT at least, quite possibly the NCAAs. St. John's, having lost over 50% of their PT from last year, and having been lucky to the tune of more than 2 wins may be due for a set back. 4-5 wins may be pretty close to where they finish. Sometimes a big recruiting class is not an advantage.

No comments: