2017-18 quick observations

[shadow]

Clippers have third worst luck rating in the league by my calculations. I think they'll turn things around a bit.

[Crow]

Thanks. Probably. I don't expect any change in next few weeks. But Ballmer could use recent losses as excuse if he really wanted to change. Is he open to extension? I dunno.

[Crow]

H Barnes on court, Mavs lose by 15 pts per 100 possessions. Barnes off, Mavs win by 10.

Star role, replacement level impact on prominent player metrics.

If the plan is still build around him...

And Smith? He is also bottom 5 in league for position on RPM. Even worse than Barnes on boxscore metrics. More reason for patience but not as easy as some said.

[Crow]

Also in bottom 5 on RPM by position: Mudiay, Rose, Muhammad, Saric, Maker.

[Crow]

Blake Griffin stable to down on almost everything. Way up on 3pt rate, slightly up on 3pt fg%. 71st on RPM estimate at a modest plus 1.

[Crow]

Giannis had a modest plus minus surge but his on the court plus minus is now back to just a fraction better than neutral. You got to improve that, a lot.

Bledsoe great so far with Giannis, Middleton and Henson. Meh with Snell and Brogdon overall, so probably terrible with them when not with most of the first 3. Horrendous with Maker (no surprise, unless you are an eye test / young guy hype hoper).



Cavs' rebound is entirely because of defensive improvement from terrible to average.

Aaron Gordon's 3pt fg% is still great for season but has gone from super high early to below league average for last 10 games. Has degree of contest / closeness of contest changed? Probably.

Wolves defense still totally unacceptable with Towns on court. RPM estimate down from plus 2.1 to just 0.8 this season.

Ballmer totally shook off / resented the Doc in jeopardy talk. Of course that is normal and could change at some point.

[Crow]

Saric's box score stats trending up.

[sideshowbob]

Here's pure RAPM for 2017-18 so far:

https://docs.google.com/spreadsheets/d/ ... ingle=true

LeBron is 4th in ORAPM and 435th (3rd worst) in DRAPM.

I'll update it periodically throughout the season, probably once every few weeks.

When you say "pure RAPM", can we take this to mean single-year, no-prior (zero prior)?

Also, could you clarify what is meant by "Total RAPM"?

[Crow]

"Total RAPM" is ORAPM - DRAPM by formula it appears; but I don't know what that means.

[Crow]

Evan Turner is still Evan Turner and was & is a bad signing. 9th lowest SG on RPM estimate.

[sideshowbob]

"Total RAPM" is ORAPM - DRAPM by formula it appears; but I don't know what that means.

Thanks!

Yes, that is what I figured but I am unsure why that needs to be represented.

Do we have an idea of how often shadow is updating this file?

[shadow]

When you say "pure RAPM", can we take this to mean single-year, no-prior (zero prior)?

Also, could you clarify what is meant by "Total RAPM"?

Yes, it's single-year, zero prior. I guess the term most people used in the past was 'vanilla RAPM'.

Total RAPM is just ORAPM - DRAPM to identify who are the best over or under players when it comes to projecting total points scored. LeBron ranks number one in Total RAPM at the moment as the best 'over' player because he's 5th in offense and 443rd on defense.

I probably will only update the file once every few weeks. I'll add a field indicating the date of the most recent update.

[sideshowbob]

When you say "pure RAPM", can we take this to mean single-year, no-prior (zero prior)?

Also, could you clarify what is meant by "Total RAPM"?

Yes, it's single-year, zero prior. I guess the term most people used in the past was 'vanilla RAPM'.

Total RAPM is just ORAPM - DRAPM to identify who are the best over or under players when it comes to projecting total points scored. LeBron ranks number one in Total RAPM at the moment as the best 'over' player because he's 5th in offense and 443rd on defense.

Thanks!

I probably will only update the file once every few weeks. I'll add a field indicating the date of the most recent update.

That would be great! Always cool to have this available! I (and I am sure many others) really appreciate it!

[shadow]

Sure, no problem. It's not that it's a huge effort to calculate it more often, but the numbers are not going to change a lot on a daily basis, so I figured every couple weeks should be sufficient.

[Crow]

Other RAPM split possibilities I hadn't considered or mentioned before (though like almost all the others they probably won't be done):

RAPM for possessions where player used the possession vs not
Touched the ball vs not
By # of player touches that possession
By # of team touches that possession
By next to last touch (or the one before that)
By degree possession time dominated by primary ballhandler or possession user
By dribble data
Player went in paint vs not
Ball went in paint vs not
Stated stint vs not
By shot clock part
By degree of spacing at end of possession or end -1 or 2
By side of court at end of possession
By # of side to side shifts in possession
By play type (including fast break and off offensive rebound and out of time out)
By distance from basket at end of possession
By sum of lineup usage splits
By sum of lineup assist % splits
By height, weight, age / experience of defender
By quality of your PG on the court
By length of court stint
Regular vs back to back games
By left or right-handedness of defended and by your PGs
By height, weight, experience or speed of lineup or the net for lineup matchup

Yes it would often require player tracking data. Yes it might need to be multi-year.

Team and league trends may be important to consider, maybe more so than player level.

Honk (publicly or privately) if you have thought or heard of (or like) any of these efforts to better understand context & chemistry or play design & process.

In lieu of the RAPM splits, the raw data would still be worth seeing. But would rather adjust I think.

Quality of the estimates is worth keeping in mind but opening up research topics and getting leads to follow & guess about are important too. Better descriptive understanding of the world is useful even before you try to evaluate & impact it.

Don't like or trust a "black box"? Then work like heck to try to understand what you can about every factor, every dimension, every aspect of the production process. The surface, guts, context & connections.