APBRmetrics

[deepak]

Discussion between Daryl Morey and Bill Simmons:

(EDIT: fixed link)

http://espn.go.com/espnradio/player?rd=1#/podcenter/?id=4936623&callsign=ESPNRADIO&autoplay=1

They touched on a lot of topics, including the limits of adjusted +/- (around the 28 minute mark), lineup analysis, and the upcoming Sloan Conference.

[schtevie]

I think this is the link:

http://espn.go.com/espnradio/player?rd=1#/podcenter/?id=4936623&callsign=ESPNRADIO&autoplay=1

What I got on the other is Hollinger.

Simmons remains a willful dope on APM. I think if it only showed Tony Allen to be a stiff, he would figure things out.

But I learned that the Sloan Conference is sold out. Dang. If beer will be drunk afterwards, do tell the location.

[deepak]

[Qscience]

Lol Bill and Daryl rips the adjusted +/- minus system...

GJ Bill asking so many +/- questions...

Totally wish I couldve made that March 7th, MIT SLOAN conference plans earlier but sigh...

[deepak]

[Crow]

If anyone were willing to offer any further detail about what was said about the limits of adjusted +/- or the value of or how they do lineup analysis- - if it goes beyond what if generally well-known- I would be appreciative. (Right now my computer's audio capability is down.)

[Qscience]

Well basically he went into how you cannot trust it because one year the one of the Collins twins was the best player in the NBA then basically goes on to say that even though Kevin Martin is consistently one of the worst +/- players they through it out because they feel that +/- minus doesnt show the high caliber player that they believe K.Martin is.

Formulas no lineups no he said something that all +/- does it tell a story of the game but if he uses it somehow other than that he isnt telling cause he eluded to public domain as places to find info and that info out there is growing.

[Crow]

Alright, thanks.

Since they won't say much, understandably, left to guessing on methods... and looking at results.

Of the Rockets with Kevin Martin and other stuff.

[ARucker]

Very surprising that Morey would point out the limitations of APM immediately following a decision to acquire a player who rates poorly in that stat. /sarcasm

[deepak]

What I took from it is that +/- stats may tell you what happened, but not why it happened. If a player you like has a poor +/-, then you look closer at the film to see why his team struggled with him on the floor. Maybe he didn't have the right type of players around him, or the team wasn't best utilizing his skills. It looks like this is what the Rockets did with Kevin Martin.

Another example Morey mentioned was specialists like Chuck Hayes. The Rockets, internally, are probably very appreciative of all the great things Hayes can do, but if +/- says he's one of the best players in the league how much stock can you really put into that? It might mean that the coach is putting him in games situationally where he can be of most help to the team. But if he's asked to play a larger role, or extended minutes, the impact will go down quite a lot (or may even become negative).

Limited players like Hayes which teams probably don't scout for may be more effective in short bursts. That could give his team a handful of possessions where they can take advantage of his strengths, before the opponent figures out how to take advantage of his weaknesses.

[schtevie]

Not that it matters much, but which Collins, which year, and to whose APM calculation were they referring?

Edit: I see that David Lewin had Jason Collins at 7.10 in 2004-05, but this wasn't even Top 20 stuff. Is this the (mis)information?

[bchaikin]

not sure about jason or jarron collins but i first heard of +/- in the 02-03 season when winston/sagarin were doing stuff for the mavericks. in this article:

http://www.nytimes.com/2003/04/27/sports/pro-basketball-mavericks-new-math-may-be-an-added-edge.html?pagewanted=all

winston was quoted as saying "...he has no stats, but we have him (eduardo najera) as great..." and there were other articles that said he was one of the best players in the league (not sure if that was a direct quote or not)...

that season najera played limited minutes, just 48 games and 23 min/g (1103 total minutes), but shot a high ScFG% (57.9%) with a very low rate of turnovers per touch for a PF (just 4%). so he was quite efficient on offense that season despite being very low touches/min on offense...

[Crow]

Thanks deepak_e.


Rollup Adjusted +/- does not tell you specifically why the team impact with the player on the floor happened.

But, yes, you can look closer at the film... and Statistical +/- and Adjusted +/- Factors and everything else... to try to see why.


With Kevin Martin in Sacramento the multi-season Adjusted Factors say that he had a very positive impact on getting points from the free throw line and team eFG% but that his offensive impact was still more than offset by his defensive liability with 3 of his defensive Factor ratings being in the bottom 10% in the league.

The defensive factors for Chuck Hayes has him in the top 5% on three.

Now there is an error term that makes all these ranking levels tentative. But I tend to assume they are generally mostly right... for that context, unless there is something that makes it really hard to believe.

Will the Factor pattern and the overall Adjusted +/- follow Martin to Houston or can their context (with Yao back, not now) mitigate his defensive weakness? Ideally I'd try to find similar players who changed teams and then check multi-season Adjusted and its Factors and see how much these marks persist or vary. We don't have the exact same method multi-season Adjusted and its Factors in public for before and after movement to allow this better comparison (better than comparing single season roll-up Adjusted +/-) but we could, eventually.

Rather than talk about the limitations of roll-up Adjusted and continuing importance of context and end it there you could also compute roll-up Adjusted +/- and Adjusted factors for Martin with Yao or Hayes vs without or with any other pair or lineup building block or type to see what further insights that gives you on Martin, pair interaction and the metric. And compare the messages of Adjusted +/- and its factors with Statistical measures and perhaps combine the input. With the talk about the limitations of pure, roll-up Adjusted +/-, especially the one season version by the traditional method you can wonder about how hard and far they've pushed on, not past the limitations but in spite of them. Wonder, but can't 'know".

Does Morey use his own version of regularized, multi-season Adjusted +/- down to the Factors? I would think so. How many other team analysts actively use it? Hard to guess but it has sounded like some may not much or at all.

Has Simmons actually looked at that version at all? Maybe not.


Everything is dependent to some variable extent on the lineups guys play with and the proportions of favorable and unfavorable= whether they can fully do what they do best there or are hampered and whether the context enhances or diminishes secondary strengths or weaknesses.

In the end it comes down to interpretation and judgment based on everything, but at this point I'd still use Adjusted +/- data as one input.

The season standings for the Kings after Martin became a top or the top offensive option aren't encouraging but we'll see what happens in Houston with Yao. Either Martin mainly with the starting lineup with Brooks or off the bench; as option 1B, 2 or 3. Team performance with and without Brooks will be something to watch, but try both and see.

The move was made (instead of whatever other available options now or after McGrady's contract expired) and he is likely a Rocket for a good while, whether it works well or not. Will there be appearances in the second round of the playoffs and beyond or not? That is the ultimate test and measure to me.

[jjstats]

FWIW, in his book, Winston actually singles out Martin as the poster boy for a player that racks up box score stats while consistently putting up negative +/- numbers.

[Crow]