"APBRY" Analytics Award?

[Mike G]

Some franchises certainly seem to outperform their expected wins year in and year out, while others do the opposite.
Last time I looked at it systematically (it's been a few years), the Lakers had regularly won more games each year than they "should have", relative to their point differential.
I figured it was referee bias -- a team with more fans "should" win the close games.

I'd like to see an updated study, pre- and post-Donaghy scandal.


OK, a quick check of the Lakers confirms my memory.
For 9 years, 1980 to 1988, they had 3 to 6 more wins than 'expected'. For the Magic Johnson era, 1980-91, they averaged +3.2

For the next 13 years, to the end of the Shaq era in 2004, they averaged +2.6 wins per year.
Since then, +1.0 over the last 8 seasons.

Late in close games, if a team should get the benefit of 1 or 2 extra "questionable calls", they might win 2/3 of those games.

[dtjmcauliffe]

Shouldn't we be evaluating this in terms of how well a team has done relative to their constraints. Basically, adjust for the picks the team has had, the capspace they could use, some role for attractiveness of market(New York versus Milwaukee matters) and performance relative to expectations that way.

I'd guess SA would come out on top in that, given the low picks since Duncan, and the winning they've done in the regular season.

If you're just looking at analytic work though, do we want to cancel out effects of good medical staffs and coaching, or is the importance of those highlighted by Analytics? I don't know of a way to answer that.

[DSMok1]

I would say that the true team measure should be:

Wins/total salary spent.

Period.

[Crow]

Wins/total salary spent is certainly a useful additional measure to consider; but in a league without a hard cap, it is not an sufficient answer totally on its own.

[DSMok1]

Wins/total salary spent is certainly a useful additional measure to consider; but in a league without a hard cap, it is not an sufficient answer totally on its own.

It's all about efficiency, isn't it? If I spend my first $50 million better than you do, and you spend $20 million more, and the teams perform similarly, I will have done better no matter the cap and spending total for each team.

[Crow]

You cite one example out of several possibilities. One team can be more dollar efficient than another but get beat by them by the other team's combined force of somewhat lower efficiency * sufficiently higher total spending.

[Mike G]

I thought it was all about winning titles.

[DSMok1]

I thought it was all about winning titles.

Yeah, pretty much, but that's a small sample size to evaluate franchises based on. Obviously the Lakers and Spurs have done well...

[Crow]

It is all about winning titles for organizations. I just don't think it necessarily follows that a title means great analytics produced, used and significantly contributing to the title. Generally it is probably a significant contributor but the amount may vary from tittle winner to title winner so I went looking for other approaches that try / might get at the specific analytic contribution in other ways.

[DSMok1]

Here's a table of the relevant information, from the 2000 season through the 2012 season:

Code: Select all

Tm   Avg Win% Avg Payroll Efficiency Won Conf Title
SAS    0.703      97.8%      0.718      3      3
OKC    0.568      88.5%      0.642      1
CHH    0.565      89.7%      0.630
UTA    0.561      93.3%      0.601
DET    0.549      91.7%      0.599      2      1
PHO    0.587      100.7%     0.583
LAL    0.652      116.6%     0.559      7      5
DEN    0.520      95.0%      0.547
NOH    0.484      89.5%      0.541
ORL    0.536      99.2%      0.540      1
SEA    0.473      88.6%      0.534
HOU    0.532      99.7%      0.533
CHI    0.452      85.8%      0.527
IND    0.526      100.6%     0.523      1 
MIA    0.544      104.3%     0.522      3      2
BOS    0.548      105.1%     0.522      2      1
DAL    0.661      127.5%     0.518      2      1
SAC    0.499      96.4%      0.518
LAC    0.391      77.8%      0.502
CLE    0.471      95.2%      0.495      1
PHI    0.504      106.2%     0.474      1
MIL    0.465      99.0%      0.470
ATL    0.421      90.1%      0.467
CHA    0.352      75.8%      0.464
TOR    0.434      93.7%      0.463
MEM    0.443      97.1%      0.456
MIN    0.446      98.6%      0.452
GSW    0.384      90.6%      0.424
NJN    0.439      103.7%     0.424      2
WAS    0.385      94.7%      0.407
POR    0.522      129.2%     0.404
VAN    0.274      88.6%      0.309
NYK    0.435      148.5%     0.293

Efficiency is expected winning percentage with an average payroll.

You've got to give it to San Antonio, no doubt about it.

[Crow]

For 9 of the titles (everybody but San Antonio and Detroit) the team spent more than the average by a bigger margin than their efficiency margin over average.

[Crow]

http://wagesofwins.com/2012/09/26/evalu ... more-12005

From coaching or work that contributes to coaching.

[mystic]

http://wagesofwins.com/2012/09/26/evalu ... more-12005

From coaching or work that contributes to coaching.

Lol, a typical WOW fail all over that article. It starts with Berri not realising that a sample which contains only data in which all players have coaches can't be used to interpret the result as if he compared players with coaches vs. players without coaches. And then Arturo using WP48 as if it was ever shown that WP48 can predict lineup performances well. Amazing, because they will likely not understand either of that.

[Crow]

What? Another case where greater dialogue could potentially help? Not that I am going to make much any further efforts for that at this time. But I thought I'd at least post the link to the article.

Arturo makes no use of lineup data here. Assumed to be not relevant / unnecessary or unreliable as usual.

Flawed as it is, the analysis still says something about team choices. 7 teams at 36% correlation or above last season. 3 made the conference finals. Not the Spurs. 5 of the 7 made the second round. Not Atlanta or the lowly Timberwolves with by far the highest correlation. Chicago next to last? That seems worth noting and investigating by other means. Perhaps it was affected in part by Rose's absence due to injury, uncontrolled or discussed in the analysis? The Heat fell 9 spots on ranking year to year but got the title anyways? Doesn't say a lot about something. The competition or the method and its significance or all and maybe more. 12 year average, the champs ranked 10th best on this measure. Only the 2007 Spurs were not above average. Lakers title winners dominated as the title-winning teams with the highest correlations along with the 2006 Heat and 2003 and 2005 Spurs. Did Pop lose his way on assigning minutes? It has bounced around lately from high correlation to medium to low, more than it used to.

[mystic]

Flawed as it is, the analysis still says something about team choices.

Yeah, the teams with the higher scoring margin are usually the better teams. Now who would have thought that?

The matter of fact is, a metric, which is bad in terms of predicting lineup performances (or anything else out of sample for that matter), because it is actually is not really evaluating players, but rather trying to scale down team performance formulas to players, will not be a good help to determine the best lineups, and thus, is not a good tool to evaluate coaching. Let alone that some teams are rather deciding to play worse players in order to get better chances in the lottery. Oh well ...

Those pseudo-academics are completely unable to account for any kind of biases. There are not able to detect selection biases, it is really a shame. They are doing that for years and they are wasting their own and the readers time with that, while in reality some of the stuff is not that hard to fix.