Prior-Informed RAPM (Updated 2-3 Times/Wk)

[Crow]

Dwight Howard is estimated as one of the best on defensive RAPM but has one of the worst offensive RAPMs amongst the best 50 on overall RAPM. James Harden is the reverse, one of the best on offensive RAPM but one of the worst on defensive RAPM amongst the best 50 on overall RAPM. It is common for the players best on overall RAPM to be better on one side of the court than the other but among the top 50 there are only 6 instances of a negative rating on one or the other. When I checked JE’s current RAPM a few days ago, there were only 2 players who were +2 or better on both sides of the court. Patrick Beverley and LaMarcus Aldridge.

I wonder what the RAPM estimate is for the pair of Harden-Howard. Is it less than +4.7 (the sum of their individual estimates)? What are the offensive / defensive splits? Which holds up more their positives or the negatives? By combined individual RAPM it appears there are about a dozen teams with a better pair than Houston. Both Harden and Howard’s RAPM estimates are less than half as good as last season. The methodology has changed but this is not a good sign for the strategy of paring these two guys. But perhaps it is ok given the dramatic rise in credit share to Beverley.

Terrence Jones is at –2.4. Parsons at +0.7. Lin at –0.2. How much are they worth paying / keeping? How much is Beverley worth? Will they have to send out one of these with Asik to move him? How much are they willing to pay overall? Is this the Rocket high water mark or will changes make them better? They have up til summer 2015 to sort it out but it would be better to do some refining before next season.

[J.E.]

I wonder what the RAPM estimate is for the pair of Harden-Howard.

Raw +/- pair numbers show this pair as one of Houston's better pairs (of pairs with >1800 possessions), with an off.eff. of 111 and def.eff of 105, and I would assume that once you adjust for opponents it would rise in the rankings. It's not elite, but decent at least

[colts18]

I don't have the old RAPM files with me but by JE's current RAPM method, Durant was -3.2 in first year at SG under PJ Carelismo then improved to -1.5 the 2nd year.

It is better to judge a player by 2nd or 3rd year than first but it seemed like a question that would be worth checking (along with others) if there was enough time and interest.

Here is Durant in 2008. -4.2

http://web.archive.org/web/201110040114 ... /ranking08

In 2009. -4.6. Bottom 12 alongside with guys like Javale McGee

http://web.archive.org/web/201203191210 ... /ranking09

[Statman]

I don't have the old RAPM files with me but by JE's current RAPM method, Durant was -3.2 in first year at SG under PJ Carelismo then improved to -1.5 the 2nd year.

It is better to judge a player by 2nd or 3rd year than first but it seemed like a question that would be worth checking (along with others) if there was enough time and interest.

Here is Durant in 2008. -4.2

http://web.archive.org/web/201110040114 ... /ranking08

In 2009. -4.6. Bottom 12 alongside with guys like Javale McGee

http://web.archive.org/web/201203191210 ... /ranking09

One reason why RAPM drives me crazy a bit - different results from different people - OR the results just seem to change with time. I never know what to fully trust, since RAPM has such outlier results that make me have doubts about the application.

It'd be so much better for RAPM if the NBA season was 1000 games.

[Bobbofitos]

It'd be so much better for RAPM if the NBA season was 1000 games.

I dunno, those 3/day games are really tough on the body.

[Crow]

Thanks for providing the old data colts18.

Yeah RAPM is still a very rough tool, best for finding most of the top third or quarter of players and most of the bottom but not all and not enough to do much definition within the middle range.

I still wonder if there would be more reliable results for a blend, of different versions of RAPM and / or SPM.

[Crow]

Anthony Davis currently -1.2 on overall 2013-4 pure RAPM and he was -1.6 last week. Almost entirely because of a negative defensive impact rating. But he will get votes for Defensive player of the year.

DeAndre Jordan will get far more votes for that award this year but his defensive impact estimate here is only modestly positive (+0.5) and was worse last week.

By this RAPM estimate the only player close to the defensive impact of Iggy (+5.2) are KG and Larry Sanders.

[Crow]

Noah had less than a +1 estimate 7-10 days ago. Now +1.3. Alright but not as good as one might haveexpected from "the eye test" / current praise.

Even though errors in estimates of smaller samples are an issue I still think various splits of RAPM would be useful, including separate RAPM estimates of players in wins and losses. Treatment themas two different players and see how much their impact isinbothsituations. Players with similar overall RAPM estimates could be very different mixes on these splits and that seems worth knowing as much as one can to add to all the other available data & splits.

I wonder how Noah looks on these splits.

[Crow]

Knicks have the most players (on team now or last with them) in the bottom 15% on RAPM estimate at 6. Other teams with 4-5: Chicago, Cleveland, LA Lakers, Charlotte, Denver, OKC, Philly and Portland. Knicks and Cavs have consultants who produce some form of APM but these teams are still on this list. I haven’t checked how many players were near bottom on RAPM last season but probably most were.

Popovich and Spurs very happy with Belinelli. But by RAPM estimate he is estimated at –3 and is bottom 15% of the league.

Presti in the last few years made moves for and got 7 guys currently in the bottom 15% in the league on RAPM estimate: Ivey, Liggins, Adams, Butler, Maynor. Thompson, Mullens and tried to get Dorell Wright also in the bottom 15%. He got rid of 5 of them but he still went for them and not others. Presti went for Butler very recently and he is currently estimated at –4 after being –3 last season. From this much, it appears that Presti doesn’t not rely on RAPM and he may not even look at it closely. Most or nearly all GMs it seems to me do not give it much weight and most may not look at it closely.

[J.E.]

Added xRAPM to my site again. If you're in doubt which ratings on my site you should use, use these

The prior is 75% my own SPM, 25% from aging (not entirely sure if that's the right split, but should be close)

The "RAPM" part is adjusted for coaches and for the 'effect of being up X points'

[Jinxed]

Added xRAPM to my site again. If you're in doubt which ratings on my site you should use, use these

The prior is 75% my own SPM, 25% from aging (not entirely sure if that's the right split, but should be close)

The "RAPM" part is adjusted for coaches and for the 'effect of being up X points'

Could you add dates to when your numbers were last updated? That would be really helpful. And thanks for all the good work!