Coach RAPM

[J.E.]

So if I am getting the explanation straight, Brooks' first-place rating is entirely (?) attributed to the above-expected performances of Maynor and Durant

No, that's very far off

[Crow]

Happened to check how the ESPN coach ratings fell on "white" and "black" coaches. White coaches make up 13 of top 15. Only 6 black coaches. Half as likely a rate to be in top 15 on ratings. 4 of 6 though were estimated positive by JE's RAPM analysis. Obviously perception and reality may not be the same and quality of coaching opportunity may not be equally distributed. Teams are also indifferent parts of cycles. Likely easier to be well regarded if team is talented. Brad Stevens rated 11th on ESPN and RAPM. 4th worst record though and actual wins 3 games under expected wins. Time will tell more.

[J.E.]

Coach ratings have been updated with all data from the regular season. Hornacek almost made it to #1. The Suns were expected before the season to go 20-62, consistent with a -7 point differential. Their final SRS is +3 for a 10 point swing. Last season they were 29th in offensive rating, this season they're 8th. That's probably not all on Hornacek - Frye coming back and Beasley going away and other moves probably helped - but he probably deserves some of the credit

[schtevie]

Yes of course Jeff Hornacek deserves some of the credit, in fact, in a very particular sense he (and the front office that facilitated his work with complementary player acquisition) deserves essentially all of it. Just about all of the year on year improvement in the Suns performance resulted from the offense (if one acknowledges that there are some minor, positive spillover effects of O on D) and I am supposing that pretty much all of that can be attributed to the consequences of a basic coaching decision: an aggressive reorientation of the role of three-point shooting in the offense, taking 8 more 3PA per 100 possessions, what led not only to realizing the long-standing, average premium over 2PA but also led to better quality 3PAs for the multiple shooters.

So, very hearty congratulations to Jeff Hornacek (especially noting the bitter irony as to how much his playing career was underappreciated owing to a coach that didn't realize his - and his teammates' - potential) but so what? Does this mean that in his second year as head coach he will also "add" +4.1 to the offense? I wouldn't take that bet. Such a gain is fundamentally a one-off, now to be eroded by laggard coaches being more confident in mimicking such an approach and bringing up the league 3PA average.

And this example clearly highlights what is a more general problem, how the perishability and "replicatability" of coaching "innovations" confuses the interpretation of +/- coaches ratings.

[Crow]

Kidd has the lowest coaching RAPM estimate amongst active coaches...
and yet, his team has the biggest actual win total for season over expected wins (+6).

I am not sure what to make of this. Yes it is complicated but this seemed worth noting for possible further investigation / discussion. I assume there is no prior for rookie head coaches like there is for players.

Along with this observation I happened to check the RAPM estimates for current coaches, who were NBA players, late in their career (but before major fall-off). The check was only for one year but of those Kidd had the best player RAPM estimate by far, with Hornachek second and Brooks close to bottom.

The only other coaches who were better than+3 were with PHI, UTH and MEM. -3 or worse were coaches with GSW, BOS, MIL, SAC, ORL and MIN with Adelman at -8 (bigger than any other negative shortfall I can recall seeing). I checked the clutch time win%s of the main Timberwolves and they were all around 40%. A little low but at least not on its face the clearcut explanation of this actual - expected win shortfall.

[J.E.]

Kidd has the lowest coaching RAPM estimate amongst active coaches...
and yet, his team has the biggest actual win total for season over expected wins (+6).

The Nets may be 6 wins over in terms of Pythagorean Wins but they're 9 wins below what the linemakers thought they'd win. Also, RAPM (obviously) ignores W/L and only cares about point differential (which also happens to be a better predictor of W/L% than W/L% itself)

[Crow]

Good points. But is Coach Kidd well above average in clutch-time as he reputedly was as a player? Have you looked into that cut of the data yet?

Does anyone have a data set (that they are willing to discuss) compares coaching behavior late in games, when up / down by more than 10 to 15+? Does Kidd let deficits balloon at the end when he feels they are going to lose anyways or let big leads dwindle more than typical (and not be scared or care) and hurt his point differential? I know some have studied this closely to aid their betting. How much is really reliable behavior vs random?

[Crow]

Possible coaching RAPM splits include not only clutch-time or factor level splits but splits for home /away, by playoff level teams vs not, by division, by offensive or defensive ratings above / below certain levels, by aspects of opponent offensive style (shot distribution above/ below certain level for fastbreak, half-court inside, 3pt.), etc. Because of estimate errors it would only be advisory but it might given some evidence based clues not identified or pursued by other means.

[TheSpiceWeasel]

(if one acknowledges that there are some minor, positive spillover effects of O on D)

Yes yes, it's only very minor. After all, the game clock stops and both teams reset after every change of possession, and the notion of "fast break" or "transition" baskets does not exist on this planet.

[Crow]

If there was coaching RAPM at the position and factor levels, one might be aided in speculation about how a coach would impact another team. The RAPM data over the years would have some player turnover to allow before and after looks that could help at guesses on coach specific impacts but admittedly in would be less than a fully satisfactory guide. Still to a curious and yet cautious analyst it might prompt deeper thought and better guesses than achieved without using this path. Won't know til it is tried.

Knowing coaching RAPM at position and factor level could help wit game strategy in general and especially that part aimed at the opposing coach and his decision-making.

[J.E.]

Coach RAPM suffers enough from noisy data as it is; splitting the data into 5 different groups/positions is only going to make this worse.

[Crow]

Just noticed VJL's spreadsheet on nba performance actual vs player expectations sorted by college coach. Will have to look at it more. Any interest in adding NBA RPM data to that database?

[J.E.]

Just updated Coach RAPM

I don't really understand why the Bucks are making a run for Jason Kidd (2nd worst rating among active coaches). You really don't need Coach RAPM to know that he wasn't great in BRK.

I'm sure we'll see the Bucks get a 1-5 pick again next season, which is probably best for them longterm, considering how bad the roster is

[Mike G]

Didn't Kidd/ the Nets actually get a lot better as the season went along?
Without arguably his best player (Lopez), they surged to a comfortable playoff spot; this after a wretched start.

Through 31 games, they were 10-21 and -5.6 ppg.
After that, they went 33-11 and were +2.9 ppg -- an improvement of 8.5 ppg.

If Kidd's 'coaching rookie' time was just 31 games, maybe he doesn't deserve his full-season RAPM?
With playoffs, that's 1/3 of his first year; likely the last 2/3 is more indicative.

[roland_beech]

Really? You put together a rating system on coaches that, oh, has Brian Shaw ahead of Pops...Rick Carlisle as a negative....and Mike B as very very negative (only took an injury ravaged team to the playoffs, and almost knocked off a #1 seed...), JKidd as very very negative (only overcame terrible start and injuries to knock off the Raptors in the first round despite no HCA...) and you feel comfortable acting as though your ratings are *authoritative* ???

Part of the reason "analytics" gets a bad name and has the problems it does getting past the initial resistance is when people fall in love with crappy ratings systems they have devised and are blind to the flaws and weaknesses. You're really going to take a stance that your 'coach ratings' should be viewed as the best way to assess coaches?