Shots at RPM

[Crow]

Happened to just notice some shots Wayne Winston took at particular player estimates in April. http://waynewinston.com/wordpress/ Probably to be expected but any analysis / rebuttal?

[BigLeagueInsights]

No idea about Babbitt, but Bogut has graded out this year as the top rim protector and one of the best rebounders (see here for more: http://analyticsgame.com/nba/stat-explo ... -data.html). He's also known as one of the best screen setters in the league.

My guess is that last year, Mark Jackson used David Lee at Center when the other team went small, and that GSW's "small" lineup was dominant, resulting in some weird +/- results. Of course, if GSW's opponent had a legit center who can score in the post, David Lee gets massacred. I don't have the lineup data to check this right now, but that's my guess.

[sndesai1]

his take on bogut is enough to make me ignore his take on rpm in general

[AcrossTheCourt]

ESPN’s RPM has Bogut well above average. I just do not buy this. Look at following

AI, Curry and Thompson on court with Bogut: 1174 minutes GSW plays 14 ppg better than average
AI, Curry, Thompson in Bogut out”" 512 min GSW plays at +23 level!
Curry Thompson in AI out Bogut in: 424 min GSW plays at +3 level
Curry Thompson in, AI and Bogut out GSW plays at +8 level
In the immortal words of Arnold of the Book of Mormon, I hope someone at ESPN will “Man Up” and explain how Bogut is above average (Our Adjusted + – has him below average)

I'm guessing Draymond Green is the missing piece here.

Seems to love Babbitt but not Bogut. That seems ... odd.

He's doing a lot of simple in/out comparisons, and while there are useful and interesting sometimes they can be misleading and it's why we do adjustments using every single player like with APM.

I'd say the priors are the differences. Jordan was rated really high by JE's SPM from the last time I saw it. I think he uses height as a variable and Jordan grabs a lot of rebounds/blocks. His rim protection stats seem really good though.

In a two year APM Jordan rates higher than Griffin:
http://www.gotbuckets.com/statistics/apm/2014-apm/

So I don't think you can entirely dismiss the results. There are also no high minute lineups where Jordan didn't play except for like one lineup of entirely bench players. It's hard to find good apples-to-apples comparisons.

[Crow]

Team results at clutch time look better for Griffin. Winston emphasizes that in his version of raw apm.

[xkonk]

Somewhat oddly (at least to me), RAPM had Bogut as the 3rd or 4th best guy on the team (http://stats-for-the-nba.appspot.com/teams/GSW.html) and a definite positive but much worse than everyone else on the squad in the on/off type comparison (http://stats-for-the-nba.appspot.com/s4 ... _2014.html).

[Crow]

s4pm for Bogut looked pretty good in 2012 and 2013. Prior may have had something to do with it.Those were short possession seasons.

s4pm charts completely favor Griffin over Jordan last season, the two seasons before that and disliked Jordan the two years before that as well.

Got out of the habit of checking these. Thanks for the reminder.

Maybe comparisons of different versions of APM (and SPM) should be done more often to find / consider cases of conflict. The option to blend is still there.


I am having trouble reaching / reading the analytics game link on phone but see enough to recommend reading it and will do so later.

Rapm isn't the end of the conversation ideally but it can be a useful stop.

[Crow]

So what should you look at if you want to consider the impact of two players on same team, such as Bogut and Green or Griffin and Jordan?

How about BPM and RPM for each player, S4PM for each, the 5man RAPM for top 10 lineups, the top 10 4 man RAPMs that have both, one and none, the splits for all these for good, avg, bad teams overall, on offense and defense, the four factor data for all these 4 and 5 man sets, the raw and adjusted pair data, all of that for last season too...and what else? I guess all the discrete box score and statVu data... and all the video...coaching notes about roles, goals and what was accomplished and not... Is that enough? What did I forget that should be considered?

[Crow]

Oh yeah, I forgot that you also need the lineup cumulative BPM for all those 4-5 man sets and the pair, and Neil's SPM and talking practice's and others if you can get them and other versions of RAPM for individuals and sets and then comparable information on all this for your main title rivals and information about the coach's past performance with players in similar roles and probably more. Then think about all of it for say 20 hours. That might be a good start. Then talk about it for 20 hours with 3-6 people who made a similar scale effort. Then talk to the coach, players and teammates. And wise outsiders. Then work on all the new leads. Then try to sum it up. Then update everything say in a month. And do that for all players. And major realistic acquisitions. And project it all out for next 1-4 years. And think about what you should add to this routine.

[permaximum]

RPM has unbelievable bias. What do you expect for the "explanation" part?

I'll calculate yearly RAPM of 2007-2012 (basketballvalue data, don't have any form of matchup data for other years, 2005-07 data is not correct enough so those 2 seasons are out) era for some studies and uses. You can then better see the amount of bias JE introduces with it's metric. RPM is not a valid but a reliable metric.

[v-zero]

Winston does a horrible job of dealing with sample sizes, he still uses excel to do all of his analysis, his understanding and methodology are frozen in time. It's fair to say we shouldn't worry too much about his criticism.

[bchaikin]

RPM is not a valid but a reliable metric.

can anyone explain why espn's RPM for 2013-14 ranks anthony davis so low:

http://espn.go.com/nba/statistics/rpm/_ ... position/6

he's listed just 25th among PFs with an RPM of 1.43...

[Crow]

Because 24 pfs were estimated by the method to have more impact. Win% on the court had him 6th on lotto team. Positive rpm, just not major. Probably affected by prior.

[bchaikin]

Because 24 pfs were estimated by the method to have more impact.

do proponents of RPM really believe this?...

do other variations of plus/minus - APM, RAPM, xRAPM, SPM, ASPM, BPM, et al - also show his impact to be less than that of 20+ other PFs in 2013-14?...

i ask as his stats were not just good but were incredibly good...

[Crow]

I believe what I said, that rpm estimated it that way. I haven't carefully evaluated his rank based on everything. What specific rank do you give him for last year and what was your most relied upon tool for that ranking?

Both WS/48 and BPM loved him last season. But team with him on court was way way below average on defensive efficiency. efg% allowed and both kinds of team rebounding were a bit worse with him on court vs off.

His will probably rate 2 plus pts better on rpm this yr. And that will be a decent, not necessarily dead on estimate. First and second yr rpm is probably off more often than later.

The PF position had more guys over 1.5 rpm estimate than other positions except for tie at PG.