APBRmetrics

Wouldn't it be nice if we knew if a player is raising his team's PointsPerShot due to his own shooting, or due to him raising the PPS of his teammates?
Wouldn't it be cool if we could quantify how much of a player's good/bad PPS is due to his teammates? Knowing how much easier some of the star players make it for teammates to hit their shots?

I've heard questions like these in the past
'Did Iverson make it easier for teammates to hit shots because he was taking most of the hard ones?'
'Did the presence of Shaq make it easier for teammates to hit shots because Shaq needed to be doubled a majority of the time?' (the hockey-assist argument)


Well now we know!

With small changes in the RAPM framework I can now derive estimates for 'influence on teammates PPS', and for 'what PPS would this player have if he played with average teammates'

The 'forecast' version ('12-'14, weighed for recency) is here

The 'points per 100 shots' column gives an estimate for this player's PPS when playing with average teammates. No real suprises there - those with good TS% have a high PPS (Chandler, Durant, LeBron, Korver..) and vice versa. This has regression to the mean built in due to me using ridge regression: You won't see someone who's expected to have a PPS of 200 (per 100), even if he is currently 5 for 5 from the field.
Phil Pressey is at the bottom, he's currently shooting 29% from the field, with 205 FGA (no, he does not play for the 76ers). Ish Smith is 2nd to last.. it's his first year in four tries in which he broke 40%FG. He's not shooting many 3s, either, and he has an FT% of 57%.

Leading the league in 'influence on teammates PPS' is Manu Ginobili, followed by Frye, R. Felton (?!) and L. James.
Ginobili's presence adds 3.9 points (per 100 shots) to the expected PPS of a teammate.
Bosh, Nowitzki and P. Gasol, in addition to Frye, are also highly ranked. I'm guessing it's due to their ability to stretch the floor. M. Bonner and Aldridge aren't too far behind. Enes Kanter, Tyler Hansbrough and Jordan Hill are the bottom 3. These guys aren't exactly good passers (AST% of 6, 2.3 and 5.3, respectively) and they don't stretch the floor very well, either.
Monta Ellis 5th surprised me a little. Josh Smith is ranked 8th, at least he makes others shoot better

And here it is with all data from '01-'14
Brent Barry is ranked with the highest PPS - he led the league in TS% in '02, and continued to shoot well until '09. Nikoloz Tskitishvili is last - he went 190 for 626 in his career.
Steve Nash had the best influence on others, by quite a large margin (+5.6). He's follow by Wade, Pierce, Baron Davis. Shaq is ranked 12th, Iverson is ranked 18th. The Big Dog at #14 surprised me

Enjoy

This is very interesting and largely looks like what I'd guess in most cases.

Ginobili's presence adds 3.9 points (per 100 shots) to the expected PPS of a teammate.

Does this include FT?
Looking strictly at shot attempts, it doesn't include TO possessions -- Manu is somewhat a TO machine -- so does it give us more or better info than just checking offensive RAPM ?
Ginobili may do reckless things that lead to high% shots and more TO.

Same questions would apply to" infl. on shooter (defense, less is better)"

Mike G wrote:This is very interesting and largely looks like what I'd guess in most cases.

Ginobili's presence adds 3.9 points (per 100 shots) to the expected PPS of a teammate.

Does this include FT?
Looking strictly at shot attempts, it doesn't include TO possessions -- Manu is somewhat a TO machine -- so does it give us more or better info than just checking offensive RAPM ?
Ginobili may do reckless things that lead to high% shots and more TO.

Same questions would apply to" infl. on shooter (defense, less is better)"

It does include FTs

I'd say it gives it us more info on why a player's RAPM could be good/bad. If you wanted to forecast a team's point differential/Win% you'd obviously still use (x)RAPM. At the very least, this version of 'adj. PPS' can give you an estimate on the expected change in a player's PPS when changing teams

In Vegas you can bet on things like 'scores more than 20 points in the game' etc., so I think they probably have similar algorithms that forecast change in a player's PPS/eFG/TS when another player is playing vs. being injured. Although I'm sure they have something better than this

This is REALLY awesome. Helps us articulate FG defense well.

Question though: is this PPS by the "TS%" definition?

i.e
Pts/(TmFGA + TmPossession-Ending-FTA)
or
Pts/(TmPossessions - TmTurnovers + TmOff.Rebs ) ?

EDIT: One of the reasons I ask is because TS% does not accurately portray the ratio of possessions : FGA and FTA.
That is, the FTA component of TS% correlates to "possession-ending-FTA" but the FGA component correlates to "possession-ending-FGA AND non-possession-ending-FGA"

bbstats wrote:Question though: is this PPS by the "TS%" definition?

It's
points / (FGA + (FTA - and1_FTA) / 2)

(a successful 2 pointer results in a PPS of 2, a 3-pointer results in a PPS of 3, a successful 2+and1 results in a PPS of 3)

I think it's like TS% if TS% 'knew' the exact amount of and1's for each player

Looking at impact on shooters on offense and defense combined, about half the league’s players have more than a 1 point impact positive or negative. About 15% have more than a 2 point impact. About 8.5% have more than a 3 point impact. These are impacts that no metric besides APM metrics account for fully and no other metric accounts for it specifically but rather estimates impacts thru certain box score stats.

Looking at impact on shooters on offense only, about 22% the league’s players have more than a 1 point impact positive or negative. Less than 4% have more than a 2 point impact. Less than 1% have more than a 3 point impact.

Looking at impact on shooters on defense alone, a bit less than half the league’s players have more than a 1 point impact positive or negative. About 15% have more than a 2 point impact. About 3% have more than a 3 point impact.

Much more difficult to find guys who have impact on teammates on offense. But even tougher to find guys who have large impact on either side.

Adjusted PPS impact on teammate offense estimate for Rondo is barely positive. J Green, bottom 4%. Other factors to consider but are either of these acceptable cornerstones of team that takes such a poor shot distribution?

I'm sure I probably missed this in another thread but how much is it weighted for recency?