APBRmetrics

[Ilardi]

Hi All,

Aaron Barzilai and I are pleased to announce that - at long last - we have published our latest set of adjusted plus-minus ratings for the 2007-2008 season: http://82games.com/ilardi2.htm. There are two features of the ratings that you may find of particular interest:

1) each player's offensive and defensive contributions have been modeled separately (treated by the underlying general linear model as separate variables); and

2) we have used five seasons' worth of data (weighted heavily toward 07-08 ) to greatly reduce estimation noise (standard errors). In fact, for the offensive and defensive ratings, the typical standard error of high-minutes players was down around 1.1. This, of course, makes the ratings considerably more useful than previous "high-noise" versions.

Please note that we're still in the process of revising the article itself to include more detail about the underlying methodology (and to make the sortable tables even more useful), but we wanted to get this up in preliminary form before the new season gets too far under way.

I'll close with a few observations - drawn directly from the article itself - based on the ensuing ratings:


1) Kevin Garnett was arguably the league’s most valuable player last year, with an astonishingly high net adjusted plus-minus rating of +14.47, nearly 3 points higher than that of the next highest player, Lebron James (+11.65). This difference was statistically significant (i.e., outside the margin of error).

2) Both Garnett and LeBron James had a more substantial impact than Kobe Bryant, a fact largely attributable to Bryant’s merely average defensive rating.

3) Although Chris Paul was one of the Top 3 offensive players in the league (+9.24), he was actually a liability on the defensive side of the ball (-4.54). As a result, his overall contribution (+4.69), while impressive, was not commensurate with that of a top MVP candidate.

4) Short point guards (Paul, Tony Parker, Mike Conley, T.J. Ford, Damon Stoudamire, etc.) pretty consistently show up in the model as defensive liabilities. We leave it to interested readers to provide a definitive explanation for this phenomenon. (We suspect it owes at least in part to their difficulty contending jump shots, especially when called upon in rotations to defend players that may be 6 or more inches taller.)

5) Because of his defensive prowess (+4.52), Ron Artest was rated as one of the top 20 players in the league last year. Historically, Houston has obtained many players whose impressive adjusted plus-minus ratings belie their less-heralded reputations: Artest (+6.31), Carl Landry (+6.55), Chuck Hayes (+5.55), and Shane Battier (+2.69). If Yao Ming (+4.77) and McGrady (+2.85) remain reasonably healthy this year, the Rockets may be regarded as the leading contenders to emerge from the West.

6) Due to their poor defense and (in many cases) inconsistent overall effort, several players with eye-popping boxscore stats appear to be much less valuable than widely believed, among them: Carmelo Anthony (-1.83), Al Jefferson (-4.61), Kevin Martin (-2.16), Richard Jefferson (-2.45), Ben Gordon (-3.37), Jose Calderon (-0.55), Michael Redd (+0.08 ), and Carlos Boozer (+0.02).

7) Jamario Moon, the 27-year-old rookie who cracked the Raptors’ starting lineup last season with his relentless hustle and off-the-charts athleticism, emerged as one of the league’s better overall players, with a total adjusted plus-minus rating of + 7.07. Because there was only one season’s worth of data available to gauge his performance, his standard error was on the high side (1.82) [7]; still, we can be 95% confident that his true overall value last year was at least +3.6 points [8] per 100 possessions – i.e., at worst, he was very good. It remains to be seen if this was something of a fluke, or if his performance this season will justify his elevation to the pantheon of elite NBA players.

[Mike G]

Of 74 PG's (>300 min), only 21 (28%) have overall numbers >0 . More than half are >0 on offense, but just 1/4 are beneficial on D.
As a group, PG's average adding 30 Points on offense (led by Nash's +580, Paul +579). But they average -48 on defense ("led" by Paul @ -284, Barbosa -274).

SG are led in offense by Kobe (+596), but are only +11 as a group; compared to an avg minus-28 defensive liability, they match the PG's in net-negativity ("led" by Redd and Durant).

Centers net nothing; all gain is from the forward positions.

Does a team do better when playing more forwards and fewer guards?
_________________
`
36% of all statistics are wrong

[Mountain]

Congrats and thanks Steve and Aaron for the article which probably will be considered one of the most important to date in the field.

I'd be interested in hearing more about clearly about the yearly weights if you are willing and the search to set them.

The future directions all make sense. Do your schedules suggest we will see more on these this season or down the line?


Although adjusted +/- tries very hard (the best) to account for context... but will a Jamario Moon with Jermaine O'Neal around now score similar on adjusted +/- as he did last season? Context is opportunity or lack of it on one kind or another. Season is just a few games old but Moon's rebounding has just evaporated- so far (might just be freak games). No longer in the role of #2 rebounder, at least all the time, lineups will matter. Maybe he'll come out fairly similar in the end by being a better scorer, perhaps some on the break or by the adjusted method fully accounting for the change in teammate strength and his opportunity. One test case to watch.

These ratings help make the Celtics title win easy to understand with Garnett & Pierce giving the team a +23 boost.

But James at +11.65 and no other Cav above 0 except Ben Wallace barely? Is the cast that weak or is the Cav so incredibly oriented to get good results from James first so far and unfortunately largely only? I guess judging how weak the rest of cast is vs the LeBron orientation is hamperedby how much these guys have been around him but there are some new faces to try to get at this.

I guess Cavs have changed their offensive system to some degree? http://thepaintedarea.blogspot.com/2008/11/i-am-witness-to-lebron-in-post.html
An admission the old one wasn't very good? But is the new one better? Better for LeBron but just as LeBron centric or better for others too?

More shocking the Lakers are basically the same with Bryant. How much in each case did supporting players adjusted +/- change when they played alongside these stars? Is there a general direction of their opportunity to contribute positively dying up?

From a 5 year perspective, a perspective not completely dominated by the Pistons context, Ben Wallace was not a good buy for 2 teams.

The findings on short PGs remains consistent with what I've thought and found earlier though I'd still like to see more adjusted +/- at lineup level before I settle too far into that view as it matter less where the credit goes to individual players as it does the total team result.

Review of Morey's pick of Aaron Brooks, currently at -7.5 on individual adjusted +/- (third worst in league among 75 PGs), could depend on the results at that level. No redemption for him or that pick from player win % on court where he was essentially last on team at 46%. Brooks at 2nd lightest in league and in ten shortest ending up at worst PG defender on adjusted defensive +/- doesn't surprise me but didn't stop the pick. Though I guess if you see Brooks as just an offensive microwave option available off the bench rather than the future starter and feel you are so stacked on defensive +/- that you can sustain him especially if you are down and have little more to lose then maybe it is more understandable. But surely there would be offensive threats available from PG that aren't as big a defensive liability or better balances. And Brooks is -1.3 on adjusted offensive +/- anyways so he doesn't seem like a strong call despite the flash.

Spurs brain trust not being able to tolerate Udrih any longer even the rest of the season (probably in part because he'd soon have to be paid more) harder to accept given his smack dab in the middle adjusted +/- while Jacque Vaughn was 4th worst. Vaughn stylistically fit the Spurs MO but didn't deliver on it.

Which team has or is the first to develop adjusted +/- for college players? Will that improve picks? Did Carl Landry have a college adjusted +/- that would have tipped off his 4th best at PF albeit small sample and probably a very favorable context? Would it have helped in the evaluation of Jeff Green, Al Thornton, etc.?


By the way adjusted +/- says Billups and Iverson are about equal. I wrestle with accepting that finding and really question Joe Dumars for doing this trade instead of some other player deal. Especially if this is to some large degree about clearly the path - in a year- for Stuckey (in bottom third on adjusted +/- right now and seems to me they are giving him too many shots and minutes too fast before earning them the old-fashioned Pistons way with defense though this method scores him better - average- on D than Eli's).

[deepak]

[Mountain]

Brooks as a low 1st round isn't a big deal but Splitter and
M Gasol were available to become a true center sized future Yao backup or even Sessions if you wanted a PG though the jury is still out on him too.

And if you really knew Landry and didn't largely just get lucky you could have assured yourself of him instead of letting 4 more teams or the teams they were talking to (Spurs?) have a crack at him and locked him into a cheaper for longer salary.

Or traded that pick and whatever for a PG you liked more proven, more balanced. But I guess that was going to be Francis.

I may be or considered too hard on GMs but it is because I think many others are too easy (if the standard is set high). But to each as they see it or decide to say it.

[deepak]

[Harold Almonte]

Ilardi wrote:

[Mountain]

McGrady delivers about the least positive impact of any the top 15 highest paid in the league.

The pattern for the best type of SG might be good on defense with a dual threat 3 point / inside game. McGrady just isn't as good himself or for his teammates as the very top guys at SG though he is 10th. But 10th at your position when you are making $19 million and expected to be #1 or 1a or b is not good return or a great stand pat position.

Dumars says Billups is not enough. I assume Morey is saying McGrady is? (Or just isn't as tradeable?)

[NickS]

Great Stuff, thanks.

[Mountain]

Thanks for those notes.

Ronnie Price is probably a fluke.

Amir Johnson probably isn't as high as he rated against weak competition. Adjusted accounts for opponent strength but ability against starters still to proved and might not follow the average step scale, especially for an interior player so thin at this time.

Herrmann better than his rating? Maybe when healthy.

What about for Villaneuva on a good team in a good context? I am not saying he is, just that there is still some chance. +2.6 in 05-06 in Toronto, still not an ideal spot. A possible outlier victimized by bad context or maybe just a little less bad a performer? A good first few games this season by individual boxscore stats. Maybe the context has been better adjusted for him.

[Harold Almonte]

Mountain wrote:

[bchaikin]

Both Garnett and LeBron James had a more substantial impact than Kobe Bryant, a fact largely attributable to Bryant’s merely average defensive rating.

by my numbers the lakers as a team in 07-08 had the league's 5th best defense, i.e. they were 5th in the league of 30 teams in least points allowed per team defensive possession. bryant played close to 1/6 of all of the team's total minutes. if bryant was just an average defender, who praytell was playing defense such that the team as a whole was the 5th best in the league?...

[Harold Almonte]

Guess you, the front court, defending the most of the scoring action (it happens near the paint).

[Mountain]

These long-term adjusted +/- results suggest it might be Gasol & Odom and to a lesser extent Fisher & Farmer. And Eli's numbers for just last season agrees on 3 but not Fisher. Brown, Turiaf and Bynum also chipped in more- by this method- than Bryant.

[Mountain]

Even if you gave the scorer and scored upon (by counterpart at best you can) just 1/3rd the credit and blame a priori would it fail to do better than the current agnostic to those immediately involved in the final play action adjusted method?

If it did, it would bolster the case for the current method.

But the door was left open to supplementing pure adjusted with something else again based on individual stats from standard or new analysis sources just as Dan did before.
I hope it is pursued further whatever the findings.

These results are better than high error 1 year pure adjusted +/- but it is still pure adjusted.