APBRmetrics

Based on circumstantial information or impression, does anyone have an opinion they want to express regarding which team with a recognized statistical analysis or advanced analysis staff member has done the "best job" with that analysis or "best job within their context and talent level" or whatever over the last 3-5 years? Or does anyone want to propose a way to make this more evaluation objective and rigorous?

Crow wrote:Based on circumstantial information or impression, does anyone have an opinion they want to express regarding which team with a recognized statistical analysis or advanced analysis staff member has done the "best job" with that analysis or "best job within their context and talent level" or whatever over the last 3-5 years? Or does anyone want to propose a way to make this more evaluation objective and rigorous?

How is it anyone but San Antonio?

San Antonio is certainly low key about what they do. No titles in last 5 seasons. Only outperforming expected Win% by 1 game per season over the last 5 seasons. Still I'd agree they are at least top 5. Really good on offense recently but that is only half the game.

The meaning of this measure is not cut and dried for talent, coaching or analytics and it is hard to separate them. But it seems worth checking.

Dallas is a top contender on this measure. +2.2 per season. OKC almost +1.

Lakers don't usually get considered a team with significant analytic resources or orientation.

Cavs had a +2.6 extra wins per season compared to expected wins for the last 5 years but 2 big-time shots with the best player in the game and only a conference finals loss really to show for it.

Griz might be beginning to make a case. Slightly negative for last 5 season but were +3 last season.

I don't know that any team's analytic shop has a really obvious and compelling case for major impact yet.

George Karl and the Nuggets (with analytic staff for at least the majority of that time) are ever so slightly negative on actual wins compared to expected wins for the last 5 regular seasons and that includes a -1 for last season.

Boston slightly negative on average for the last 5. Same for Miami.

The Rockets were also negative on average on this for the last 5 seasons.

Indiana even for the last 2 seasons. Milwaukee with a poor -3 each of the last 2 seasons. Toronto even over the last 3. Orlando -1.6 on average over the last 5. Phoenix +1 each of the last two. Portland slightly negative for the last 5.

Actual wins - expected wins is a kind of win efficiency measure. Good talent, coaching and analytics can enhance that or do the opposite- i.e. more points padding of wins which would depress this win efficiency measure. So there isn't a simple, clear, universal interpretation but it still seems worth knowing which teams were most win efficient on this measure and which teams were not in that top group. It is worth knowing overall what the superior production is yielding.

Mavs and Cavs were the leaders on this win efficiency measure and both have and / may use in-house versions of APM. The Griz had one of the best marks last season and also have an in-house source of APM. May just be a coincidence / perceived pattern or maybe not.

I forgot the Nets. They were +3 last season.

Warriors slightly above neutral over the last 2 years.

Detroit +3 last season after averaging -1 the previous 4 years. Related to their analytic hire from Orlando or not? Can't really say from the outside but it looks better for the analytic staff that way than for the reverse win efficiency measure. Orlando though was +2 last season after averaging a -2 the previous 4 seasons. Maybe the analytic staffer was not the difference maker in either case.

Whether it can ever be tracked accurately or not, there is probably plenty of room for greater and better use of analytics and / or greater and better development of analytics, (with or without really complicated expert-level computer languages and statistical methods).

Crow, why are you using X wins above/or below expectation as the barometer for an organization?

As I said, it is a win efficiency measure and it seems worth knowing if a team is strong of this or not. I am checking it to see what it says in the absence of much else that is even this solid or potentially useful. I asked for other approaches but no one has offer any.

A team that wins an unusually high % of games, relative to their point differential, may be ...
a) ... lucky. In an 82 game sample, and a smaller sample of close games, this could well be the biggest part of it
b) ... good at winning close games. Still, impossible to distinguish from luck.
c) ... good at coasting to wins in games that are well in hand. If you run up the score unnecessarily, you'll jack up your pt-diff, but not your win%.

There could be other indicators, but I don't see how an analytic advantage would factor in.
A coach might consult the numbers and stick with his best lineup to finish very close games?

A GM who uses APBRmetrics to build a great team (NBA's Billy Beane) is who should get the award. Not necessarily performing above or below expected wins.

Ainge built Boston with the help of McHale and Presti for 1 title. The role of analytics there is hard to gauge but it appears to be a very junior partner to the trades and the social ties and skills that helped facilitate those trades.
Kupchak built the 2 title-winning Lakers without much publicly noted and credited analytic help and with the help of David Stern. Great job though.
Mavs used analytic help but it appears they also got really hot in 2011 playoffs after many unsuccessful tries. But good return on analytic focus on clutchness, I think.
Not sure how much to credit Miami's analytic efforts last season, in part because of the low win efficiency information. Good defense produced in large part by good film analysis is an analytic contribution, but maybe not a high-powered statistical analysis contribution.

Morey in a 14 way tie for last among executives on the basis of no NBA, conference or division titles. Only the two Grunwalds and Chris Wallace have been at it longer with no banners produced.

Only 11 of 30 have ever won a conference title (the level that really means anything to me). Three "big" names (Walsh, Ferry and Presti) have only done this once. I tend to think a big name among executives should generally have more than one of these but I guess some might qualify on overall body of work.

In general, performance above or below expected wins based on point differential is mostly luck. I know in football, the year to year correlation on Wins-Expected Wins is 0.02, or basically none at all.

I would not use it to evaluate anything except how "lucky" the team was.

So you are saying luck is more important than coaching or analytic impact on win efficiency on average. The stats may back that up on average. So then what does that say about NBA coaching and analytics to date on average? Not impressive on average or at least not proven to be by this measure.

The data might still be "somewhat useful" in identifying some of the best / worst outliers on coaching and analytic impact, even if that claim can not vigorously supported / proven with statistics. I'd still rather know where teams fall on this than not.

Crow wrote:So you are saying luck is more important than coaching or analytic impact on win efficiency on average. The stats may back that up on average. So then what does that say about NBA coaching and analytics to date on average? Not impressive on average or at least not proven to be by this measure.

The data might still be "somewhat useful" in identifying some of the best / worst outliers on coaching and analytic impact, even if that claim can not vigorously supported / proven with statistics. I'd still rather know where teams fall on this than not.

When you say "win efficiency", you are referring in essence to "clutch" performance (meaning more close game wins than losses). And in general, quite a bit of research on clutch performance has been done, and it's extremely unstable/not predictable/random.

And Mark Cuban feels otherwise. Either to take credit for being lucky or for really valuing and understanding clutchness more than most other folks and getting good results on it, at least in 2010-11.

"Mark Cuban said so..." isn't the basis for a strong argument on a stats board. Cuban could be wrong. Not saying he is, but I haven't seen any evidence yet to support the idea that expected wins - actual wins is a good way to measure the effectiveness of a team's stats program. Nor have I seen evidence showing that the difference between expected wins and actual wins is based on superior clutch play or whether it's just luck.

This is an issue I'd love to study in detail, but the paying job makes its demands.