December 5, 2009

Fun With Hoop Numbers

If you've not been to Hoop Numbers, do yourself a favor and give it a look... it's fantastic. Their bag is APM -- adjusted plus-minus data -- and they adjust that shit more than most, using a statistical technique called cross-validation and three seasons' worth of data to filter out noise. Or something. (Look, you don't come here because we're stats experts. You don't come here, period. Nobody knows this blog exists.) Plus-minus numbers always feature some quirkiness, no matter how well you chop 'em up; according to HN's regularized APM, the 15th-most effective NBA player, based on '06-'09 data, is Matt Bonner. Nobody, including Hoop Numbers Head Nerd (HNHN) Joe Sill, would argue that Bonner's a great player. But it is possible that the Spurs use Bonner to great effect... he has one strong skill, particularly for his position, and in role-player minutes for an intelligent team, a one-skill guy can make a big impact. The question this measurement asks is not so much "who's the best player?" as "who helps their team the most in the role they're given?"

RAPM is like any plus-minus system: LeBron tops the rankings with a +5.973, which means he makes a team about six points better per 48 minutes. (Doesn't sound like a lot, but it is. Dallas was about six points better than us last year, outscoring their opponents by 2.0 per game while we got outscored by 3.7 per game. Those 5.7 points created a 21-game difference in our records.) Hoop Numbers also splits this figure into offensive RAPM (how much a player helps his offense) and defensive RAPM (how much a player helps his defense -- negative numbers are a good thing here). They dice up the numbers in some other fun ways too, but we'll get to those another day.

Based on weighted three-year averages of '06-'09 data, here's where all the current and former Warriors rank on the list of the most effective NBA players (current Dubs in gold). You're going to want to New Tab this bad boy.

* A bummer you'll notice quickly: Randolph (and Wright, and Kurz and others) is not listed, as he played too few minutes to qualify for credible regression analysis. More generally, the fewer minutes a guy played over the last three years, the greater margin for error in his RAPM. So extra grains of salt for guys like Webber, who barely qualified.

* On the subject of grains of salt, these numbers are weighted towards the '08-'09 season. Thusly, guys like Baron and Monta are heavily penalized for sucking last year. Both rated as top-100 guys the previous season (Baron was actually 24th), and both will probably rejoin the 100 with their bouncebacks this year. Still, they did both suck last year, so the rating's not entirely unfair.

* By HN's accounting, the most effective player to wear a Warriors uniform in recent years is... Al Harrington. OUCH. There are definitely more than 45 guys who are better than Al, but I don't think this rating is a total fluke, either. Al, while a mediocre player, is an extremely useful *type* of player, a floor-stretching four who can defend credibly. Fours who can shoot threes -- Al, Bonner, Rashard, Rasheed, even Donyell -- tend to fare very, very well by this system. That could be a flaw in the methodology, but it could also indicate that fours who can shoot from distance are extremely valuable weapons in the NBA... league trends suggest that the latter holds some water. At any rate, we did consistently well when Al was on the floor for us. He was a useful piece, and Nellie did us a disservice by alienating him. (Vlad could theoretically fill the same role for us, but his poor defensive numbers suggest it's not likely to pan out well.)

* Jeff Foster never actually played for us, as we shipped him out immediately in the red-hot Vonteego Cummings trade... I just thought he was worth including as an example of a type of guy we need. Strong post defender and rebounder, and not so skill-less that he'll kill you on offense. Jeff Foster is what we all feebly hope Chris Hunter can be.

* Among current Warriors who haven't recently undergone hotly debated wrist surgery, the highest-rated guy is Anthony Morrow, at 88th. For your highest-rated guy, that's not very highly rated... the only NBA team that didn't have at least one player rating higher than Morrow was the Grizzlies. In general, these rankings tend to confirm what we already know: we have a wealth of above-average players but no real great ones.

* Having said that, Morrow's showing (which came in a small sample, so salt away) is pretty promising. He made our offense 3.12 points better per 48... only eighteen NBA players had a bigger impact on their offenses than that, and you've heard of all of them. Morrow gave most of those points back on defense, ranking in the lowest 10% on that end, but if he can get his defense from "nightmarish" to "just plain bad", he'll be a big, big asset. And anecdotally and based on unadjusted plus-minus so far this year, it does seem like he's making gains on D.

* Even last year, Morrow's D wasn't quite as bad as that of several familiar faces; Jamal Crawford and J-Rich rated as the 12th and 13th worst defenders in basketball, and Troy Murphy was 21st. It should surprise no one that our current players and alumni are very well represented in the lower reaches of the defensive rankings. On the other end of the spectrum, Our Man Adonal was still the 24th-most effective defender in the league during this period. (Sadly, being the 17th-worst offensive player takes a lot of the shine off that.)

* Only five current Warriors rate as defensive assets, and, cruelly, four of them are on the shelf: Biedrins, George, Azubuike and Bell. The remaining Warrior with an above-zero defensive RAPM is Mikki, and, well, let's just say he's not looking like a good bet to keep that going. Stephen Jackson does not rate as a defensive asset, either... he made defenses (mostly ours, partly the Pacers') about a point worse per 48 in this period, rating about even with weak defenders like CJ and Maggette. He rates fairly well overall, though, as the system found him to be a positive offensive contributor.

* Maggette's terrible showing is interesting: his defense only rates a bit below average, but his offense doesn't rate as a positive at all. He ranked highly only a year, so this is based primarily on his first year with us. Some claim Maggette's tunnel vision is a net negative on the flow of our offense... I don't know that I believe that, but this is the first plus-minus data I've seen that could lend credence to that idea.

* Going by these figures, the Harrington/Crawford trade made us 3.1 points worse per 48 minutes, which translates to roughly a ten-win difference. That's probably on the high side, but it nicely illustrates just how stupid and counterproductive that trade was, even before you factor in that Al had the better contract. We turned a useful player at a skill position into a bad player at a position we already had well covered. That... was... bad.

* Besides Al and the always-wince-inducing Arenas, these charts don't make you think we've cast aside any gems recently. Baron comes off as awfully warty here, the Dunmurphys are no great shakes and Jamison, Pietrus, Barnes, Crawford and J-Rich all rate as below-average players. J-Rich actually rates as a downright bad player, and while that may sound blasphemous to some, it's pretty hard to find evidence of teams profiting from his presence. That O just don't balance out that D.

* Poor Josh Powell ranks dead last overall -- 387th out of 387. I have to admit, that sounds wrong to me... while not good, he seems too serviceable to be that much of a drag. But his unadjusted plus-minus numbers are always terrible even for a garbage-time guy, so he may be pulling one over on me.

Happy Staturday!

1 comment:

doug said...

Cool stuff. The eyes really deceive in basketball. Who knew Al was actually "good?" One of the Rumors™ on ESPN Insider, of which I am not an Insider, suggests that Al's about to move on from NYK for nothing more than "a bag of balls."