Today, I thought I'd try to look more coarsely at that analysis. First, I can retroactively calculate the expected point spread, based on Adj. Off. Eff. and Adj. Def. Eff. from Ken Pomeroy. Now this won't perfectly reproduce his original point spreads, since his efficiencies are based on performance throughout the season, weighted toward the present. But, it'll give an idea of the level of competition for each game the Hoyas played. Note that I'm weighted +4 for home and -4 for road. Also, I'm cheating, and applying the pace that the game was played at to calculate points, rather than efficiencies (as always, click any image to enlarge).
Things are definitely going well for Georgetown these days. Based on play so far this season, the Hoyas would only be underdogs for the Duke and Pitt road games.
Well, we can bring reality into our hypothetical world, simply by plotting the actual point differential for all the games played to date:
Strange as it is, ODU still has the best (worst?) point differential against the Hoyas this season.
Okay, now if I simply subtract the first chart from the second, we should get a measure of how well G'town played in each game, relative to where we'd expect based on today's offensive and defensive efficiencies.
Well, by this metric that ODU game is still the worst that Georgetown has played this year, but the first game of the year, against Hartford, now looks like the 2nd worst. Some of these results are intuitive (the Oregon game and the 1st Villanova game were obvious stinkers), but some of this is surprising (e.g. Ball St. and Navy).
The best game was the St. John's game (30-4, 41-9 point run), along with Vanderbilt and Michigan. In fact, since Big East play began the Hoyas are an average of 3.9 points better than the spread. And remember that my calcs are based on statistics compiled through Monday, Feb 19th. This may be pointing to a lag in Pomeroy's adjusted statistics, in that they are weighted a bit too heavily by games early in the year.
Here's a table of games to date, ranked from best to worst according to my fancy method:
Before I wrap this up, I'll point out a couple of things from the table:
- Only 1 game in February (@ Nova) isn't in the Top 10.
- The 6 worst games were at home, the 4 best on the road. This may be a good omen for tournament season.
- The 5 worst games all had paces of 60+. Is this a trend? Well that depends.
Without Vandy Game
Two things.
ReplyDelete1. In a game that's a potential blow-out, the pace isn't necessarily going to be uniform from start to finish. When you have a 22 point lead and pull the starters with four minutes to go, and the scrubs wither that down to a 14 point win, you're messing with the stats.
2. My theory of the Old Dominion game is that because it was on campus, the players were on a completely different schedule than they usually are when they have to head down to the Verizon Center, sometimes several hours before the game starts. Since there was no commute time, did the players hang out in their apartments until shortly before walking over to McDonough to suit up?
Regarding #1, I agree that the end of games can skew that stats (the Notre Dame game in particular comes to mind). But, I'd need to work through the play-by-play data by hand to remove this effect. Box score data is readily available, so that's what I'm using.
ReplyDeleteThanks for reading!
Can your post the formula for how you retroactively calculate the expected point spread, based on Adj. Off. Eff. and Adj. Def. Eff. from Ken Pomeroy?
ReplyDeleteThanks!