Monday, February 19, 2007

Plus/Minus Redux

Raw plus/minus is nice, and I think interesting, if not statistically significant, for an individual game. But step two is to calculate an on/off rating a la the work Roland Beech does on NBA players.

This kind of analysis is problematic for star college players, because many of them play nearly all of their team’s non-garbage time minutes. We need a pile of data in the off category to make the results meaningful.

It just so happened that when the Pittsburgh Post Gazette’s Ray Fittipaldo wanted to step outside the box and do a story on the plus/minus concept, it gave me an opportunity to carry out this kind of analysis on a team where the star player does get some meaningful time on the bench. Aaron Gray is widely thought of as All-American/Big East POY material, and yet he averages 29 minutes a game. Not surprisingly, on/off data shows that Gray makes a huge difference to Pitt.

With Gray on the floor, Pitt is a +17 per 40 minutes, and with Gray on the bench, Pitt is -2. So over 40 minutes, Aaron Gray is worth about 19 points to Pitt. In his typical 29-minute game, he’s worth about 14 points. Basically, if Pitt had lost Gray to a season ending injury during Midnight Madness, they’d have a sub-.500 conference record. Which makes sense when you consider that Gray’s rating dwarfs any other Pitt player. Antonio Graves came in a distant second with an on/off of +5.

Not that this is particularly shocking, but it does illustrate the value of such an analysis in showing a player’s worth to his team. And keep in mind the value we are talking about is realtive to the team. Gray has no viable backup, thus he is very valuable to the Panthers. If we ran the same kind of analysis for Ohio State, I wouldn’t be surprised if we found out that Greg Oden’s value is in the single digits, partly because his replacement, Othello Hunter, is a solid player himself.

Check out Mr. Fittipaldo’s article for the rest of the details. (My favorite part of the article is the caption on the Aaron Gray photo - “Aaron Gray and the Pitt Panthers are still a good basketball team.” It’s almost like the captioner spent the week being endlessly harassed by a bunch of Louisville fans.) The complete list of Pitt on/off data follows, through last Monday’s Louisville game.

         On    Off   Diff
Gray    +17.4  -1.7 +19.1
Graves  +13.6  +9.0  +4.6
Cook    +13.7 +10.2  +3.5
Fields  +13.0 +10.5  +2.5
Kendall +12.2 +12.6  -0.4
Ramon    +9.1 +17.2  -8.1
Benjamin +5.8 +15.5  -9.6
Biggs    +5.4 +15.8 -10.4
Young    +5.6 +16.7 -11.2

Friday, February 16, 2007

The First Poll is the Best Poll

I’ve been kind of distracted by work on the HD box score. Here’s Wednesday’s Ole Miss/LSU game, a 71-70 Rebel win in 60 (counted!) possessions. A decidedly bizarre game where Ole Miss (now leading the SEC West, kudos to Andy Kennedy) turned it over just four times but saw that advantage offset by LSU’s slightly better shooting and much better rebounding. I really preferred to run through Radford’s 114-107 win over VMI, but the Radford scorer’s table apparently gave up on tracking substitutions at some point in the second half. I don’t blame them.

I hardly pay attention to the opinion polls, but one poll I do think has merit is the very first one of the season. It’s when the games start being played that the polls go to crap. LSU is headed to the rare fate of being a preseason top 10 team, only to not make the tournament.

In “achievement/performance dichotomy” news, I’m not one for conspiracies, but I think there is one being orchestrated by Coach K. I mean, people just keep ragging on the Blue Devils. But consider that this team:
(a) arguably plays in the toughest conference in America,
(b) inarguably plays the toughest conference schedule in that conference, and then
(c) throw in a large dose of bad luck.

What you will get with any non-elite team in that scenario is a bunch of losses. Duke is not a great team, but they’re probably very good, and they will play 13 of their 16 conference games against teams that are currently deemed at large worthy by Joe Lunardi.

It won’t take much of a favorable draw for them to make the Sweet 16, and with a few breaks here and there they’ll make the Elite Eight. At which point you’ll have to hear what a wonderful job K has done with this ragtag group of six McDonald’s All-Americans. Oops, I mean castoffs that nobody else believed in.

If that scenario unfolds, I’d say the more logical explanation for the big surprise is found in (a), (b), and (c) above. Now, Duke could easily suffer the same fate as Kansas and Iowa last season - defensive minded teams that were knocked out in the first round. But they could also be like LSU or UCLA. And if the latter happens, I’m just warning you: The foundation has been laid for the biggest Coach K love-fest ever.

And finally, thanks to Dan Steinberg for his heart-rending story, er, blog post, on my life and times. I got more nice e-mail in a couple days than ever before. It goes to show that all you have to do is beg.

Wednesday, February 14, 2007

HD Box Scores

A lot of folks send in great tips on some cool box scores that appear during the season. I wish I was dutiful enough to print them all. I did get a lot of mail about Butler’s offensive assault on Cleveland State last week. A historic game indeed - Butler’s 92 points on 53 possessions was the most efficient game (1.73 PPP) in the modern era*, knocking off DePaul’s annihilation of Syracuse last season (1.65). 

It seems like a good time to reveal my latest project - the HD box score. Much like HDTV, it’s not necessary to view a box score in HD to understand what went on during a game, but it does add some resolution to your experience. I’ve got some more work to do on this, but the player data is pretty much set. The essence of the HD box score is that opportunities are tabulated along with the traditional counting stats. The only exceptions are the points and FGA categories where the “opportunities” are the team totals while the player in question in on the court. But for assists, you have the teammates’ field goals made associated with it, for offensive rebounds, you have total possible offensive rebounds, and so on. You can also see the defensive and offensive possessions a player was on the floor by looking at the steals and turnovers categories, respectively. (My possession counter still needs some work, but it’s getting there.) The best part might be that playing time is listed in minutes and seconds, just like the NBA.

So anyway, here it is for the Butler/CSU game. Pete Campbell (28 points in “about” 31 possessions) almost averaged a point per possession on his own! I’ll try to post one of these every now and then for games of note, and where the data is available, because it’s not available for a lot of games. Much like TV, most contests are not yet available in HD.

*Modern era began in 2003-04.