Thursday, April 14, 2011

On Second Thought -- Plotting Trendlines for Big East Teams (pt 1)

What Funny Shapes Can Teach Us...
I liked tj's original post on mapping Michigan State's efficiency trends during the 2008 season when I read read it years ago, mapping the trend over time could be useful. The graph itself however was visually busy; the trendlines got lost as MSU's offensive and defensive efficiencies (points per possession scored and allowed) gyrated, game-by-game. Which leads us to the real problem of measuring team efficiency over a series of games (in conference out of conference, through a tournament, etc.) -- adjusting the performance to account for the competition. The conference regular season, which offered a more balanced schedule than a composite out of conference slate of opponents, should be a better context for judging performance over time. The bloggers over at Maize-Colored Glasses Blog put together an outstanding set of alphabetically correct posts that mapped each Big Ten team's trends over the course of conference play. I was so impressed with the methodologies described in the "Polynomial-centric Recap: Part 1" that I decided to apply the method (unadjusted and adjusted) for a similar look at the Big East. I describe the techniques later in this post. From the bottom up, (alphabetically correct of course) the conference's last five teams shared some features. The conference standings/seeds for the Big East Tournament:

WLPct.
Pittsburgh1530.833
Notre Dame1440.776
Syracuse1260.667
Louisville1260.667
St. John's1260.667
Cincinnati1170.611
West Virginia1170.611
Georgetown1080.556
Connecticut990.500
Villanova990.500
Marquette990.500
Seton Hall7110.389
Rutgers5130.278
Providence4140.222
South Florida3150.167
DePaul1170.056

(16 #16) DePaul


Not an ironic commentary on DePaul's season...honest

I rechecked the numbers and calculations when the fish appeared. The unadjusted trendlines look much the same. Coach Oliver Purnell will bring the program along next season, but seeing offense and defense move in such a symmetrical way throughout the entire season suggests (perhaps) that some nights his young Blue Demon squad got it and other nights not at all. The late season "fish tail" was probably a morale-induced slump. More experienced players should make a difference.

(14 #14) Providence


It was the defense...again

Throughout the Tim Welsh and Keno Davis Eras the Friars have been a Notre Dame Lite -- good offense but little defense. This season was no exception, and Marshon Brooks alone could not carry the team night-in and night-out. Lack of progress no doubt the catalyst that led to Davis' departure, Ed Cooley has been charged to turn things around. One common trait of the teams in the fourth quartile is that the defensive efficiency trendline never falls below the team's offensive efficiency trendline. Though everyone of the four team that finished #13 through #16 won at least one game, winning was neither prevalent (all had winning percentages <0.400) nor sustained for any period of the conference season.

(13 #13) Rutgers


An end of season rally...

The unadjusted trendline suggests the Rutgers players were "getting it" -- Coach Mike Rice's system as the season progressed. Though the Scarlet Knights went 1-3 through the last four games, the defense appears to have stabilized even as their offensive efficiency improved. Note that by the 18th game (a season-ending, 74-75 loss to Providence), trendlines for offensive and defensive efficiency seem poised to cross. And the Knights went on to win their next game, 76-70 in overtime over Seton Hall in the Tuesday night round of the Big East Tournament. The trendline chart adjusted for the competition however suggests something a bit different...


...or a modest slump?

Providence and DePaul, teams that finished #14 and #16 respectively, were Rutgers' last two regular season opponents. The Knights' performance (a 1-1 record on games with margins of +3 and -1 points), coupled with the "fish tail" curve of the trendlines suggest that the team was slumping more than rallying, as the season drew to a close. The win over Seton Hall, an instate rival with whom the Scarlet Knights drew 1-1 on the season, does not refute the trend. It will be interesting to see what Coach Rice does with a squad made up largely of players he recruited.

(12 #12) Seton Hall


The trendlines are consistent with a losing record...

The trendlines converge and cross just before the end of the season (for the curious, game #14 for the Pirates was Villanova). The wave motion is similar to South Florida's (see below), but unlike the teams that finished #13 - #16, the trendline for the Pirates' offensive efficiency crossed to the north of their defensive efficiency trendline, an element the Hall held in common with the other 11 members who finished in the upper three quartiles.


Coach Willard's system was beginning to work

The Hall's last two regular season games, both wins, over St. John's and Marquette, were a huge boost for the adjusted numbers. The Johnnies were seeded #5 in the Big East Tournament and managed to draw a #6 seed in the NCAA Tournament, while Marquette finished the Big East regular season with a 9-9 record, drew a #11 seed in the NCAA Tournament and ran to the Sweet Sixteen.

(15 #15) South Florida


The wave pattern?

Having finished #9, drawing an NIT bid and losing Dominique Jones all in 2010, the Bulls were poised for a rebuilding year, and 2011 lived down to those expectations. The wave pattern (the adjusted version above is consistent with the unadjusted trendlines), which features the trendline for defensive efficiency always above the offensive efficiency trendline, by (more or less) a consistent margin, throughout the conference season. With a 3-15 conference record (10-23 overall) and a cumulative conference record of 51-77 in 4 seasons, I wonder, lacking a clear sign of improvement next season, how much longer Coach Stan Heath will pace the sidelines in Tampa.

Methods & Sources
I used Ken Pomeroy's website, specifically each Big East team's gameplan pages, as the data source for each team's offensive and defensive efficiencies for each conference game played (many thanks to Ken Pomeroy for providing this invaluable resource for the past eight seasons). I plotted the data points for offense and defense and then applied a third order polynomial to map the trend line. Those charts are captioned as "Unadjusted" in this and subsequent posts. The conference average for points per possession in 2011 (conference games only) was 1.037. I adjusted each team's offensive and defensive points per possession for each game using the formulas documented by the Maize-Colored Glasses Blog...
offensive adjustment -- (100 x points per possession)/(1-(opponents average points per possession allowed - conference-wide average points per game allowed)). Given the Big East average points per possession was 1.037, that simplified the adjustment to (100 x points per possession)/(opponents average points per possession allowed - 0.037)
defensive adjustment -- (100 x points per possession allowed)/(opponents average points per possession - 0.037)
A strong defensive effort against Notre Dame would be properly weighted versus a strong defensive effort against DePaul.

Offensive points per possession for each team (ordered by PPP Scored):

PPP
TeamScored
1Notre Dame1.130
2Pittsburgh1.105
3Marquette1.103
4Syracuse1.078
5Villanova1.077
6West Virginia1.049
7Louisville1.034
8Cincinnati1.030
9Georgetown1.026
10TSt. John's1.022
10TConnecticut1.022
12Providence1.010
13Rutgers1.009
14South Florida0.972
15TSeton Hall0.958
15TDePaul0.958

Defensive points per possession for each team (ordered by fewest points per possession allowed):

PPP
TeamAllowed
1Pittsburgh0.960
2Louisville0.967
3Cincinnati0.981
4Seton Hall0.987
5Syracuse0.989
6St. John's1.001
7TWest Virginia1.004
7TConnecticut1.004
9Georgetown1.037
10Notre Dame1.055
11Villanova1.062
12Marquette1.070
13Rutgers1.082
14South Florida1.094
15Providence1.111
16DePaul1.172


No comments: