Wednesday, April 22, 2009

Performance vs Expectations -- Post Season

All Things Considered...
I was talking to a friend over the weekend -- we had last talked just after the 1st round of the NCAAs last month. "The 'Cats make the Final Four and not a word from you?" I teased. "Well..." he replied, "It was pretty clear that any of three or four Big East teams could make the Final Four. So it was not that much of a surprise...". I forgive him for the tepid congratulations. He cheers on the Pirates, and the past few years have not (despite the new digs) been very kind. Maybe next season. The Wildcats did a very passable version of the Energizing Bunny through much of March and into the first week of April. Very satisfying (aside from the margins posted in their 2 losses during the post season run). How did the 'Cats perform in their seven post season games (I grouped both their Big East Tournament and NCAA games together for this review) versus Pomeroy's projections? The chart below illustrates...

Note wins are marked with blue circles, losses in red. The difference between performance (the margin of victory/defeat) and expectation (Pomeroy's projection) is given relative to the X axis. Performance matched expectation would be marked with a circle on the thick black line (I wanted it to stand out...). Performances that did not match expectations (negative differential) are below the X axis; those that exceeded are above. Relative performance is noted by the distance above or below the X axis, each data point coupled with its margin & labeled with the opponent.

If You Thought the Duke Game Was the High Point...
You have an ally in Ken Pomeroy. The winning margin, 23, is pretty impressive considered by itself, but according to Pomeroy's calculations Villanova was supposed to lose by 3 points. The UCLA game is a close second. Also projected to lose, this time by 1 (Pomeroy gave the Wildcats a semi-home court bump, after all the game was played in the Wachovia Center). I was most impressed with the Pittsburgh win. The Panthers have the type of team that Villanova most struggled with during the season -- a dominant center, competitive front court, combined with a good point guard. The game was essentially undecided with <5 seconds on the clock. Those who described it as an instant classic are correct.

Ironically many in the Nova Nation were disappointed with the Marquette and American University games. Yet both were closest to Pomeroy's projection. I never had the sense the outcome of the American can was in doubt, but the Wildcats were not able to pull away in the same way they did with UCLA and Duke. Among Big East Tournament games, I believe the win over the Golden Eagles, coming as it did at (literally) the last second, was the second only to the Connecticut -- Syracuse game. It foreshadowed the East Regional Final game.

Pomeroy's Model vs Performance
Pomeroy's record for accurately predicting Villanova's wins & losses was a disappointing 4-3, thankfully. Pomeroy did predict both of the Wildcat's losses & the wins over Marquette and American University. He missed on Pittsburgh, the UCLA and the Duke games. While Pomeroy does employ an "aging" technique to reduce the influence of early season results, I suspect it is not refined enough to give proper weight to late season surges. When I telescoped each team's raw offensive and defensive ratings for just the post season however, I did get projections that suggested outcomes more favorable to Villanova than Pomeroy's adjusted ratings. Pomeroy's projections did predict 4 of the games to within +/- 5 points.

No comments: