APBRmetrics

There's a new addition to my site: You can now compare two players according to their performance while having played with the exact same 4-man-units. I'll try to explain with an example.
Let's say we want to compare Dwight Howard and Ryan Anderson. We look for all 5 man units that contain either Dwight Howard or Ryan Anderson (but never both of them). For all of those 4-man-units that exist for both players we can make a direct comparison. Let's say we have

(1) Nelson-Richardson-Turkoglu-Davis-Howard and
(2) Nelson-Richardson-Turkoglu-Davis-Anderson

(1) scored an average of 1.05 points in 30 possessions, (2) scored an average of 1.1 points in 20 possessions. It would seem that the team is scoring 0.05 points more per possession with Anderson.

Obviously, there are many more compatible 4-man-units for the duo of Howard/Anderson. We might also have

(3) Duhon-Richardson-Turkoglu-Davis-Howard and
(4) Duhon-Richardson-Turkoglu-Davis-Anderson

For this example, let's say (3) scored an average of 1.1 points in 10 possessions, (4) scored an average of 1.12 points in 15 possessions

For all those compatible 4-man-units we add up the offensive efficiencies(?) (for each player seperataly), weighing the off-eff. of that unit by the minimum of possessions played with that 4-man-unit by either player.

So if we only had those two 4-man-units for the duo of Howard/Anderson, Howard would get an
( 1.05*min(20, 30) + 1.1*min(15, 20) )
/
(min(20, 30)+min(15, 20))
=1.07

while Anderson would get
( 1.1*min(20, 30) + 1.12*min(15, 20) )
/
(min(20, 30)+min(15, 20))
=1.10

From this we can say that, given the same teammates, Anderson has had a better impact on team offense than Dwight Howard

You can reach the analysis for each player by clicking the "Info" button on http://stats-for-the-nba.appspot.com/ranking_rec
Older seasons are still being computed and should be online in a few hours

There is no adjustment for strength of opponent as of yet

A couple of notes:
Ryan Anderson is absolutely destroying Glen Davis, and even looks better than Dwight
Nick Collison is all positive in the last two years, no matter who he is getting compared with
Udoh looks better than David Lee (even on offense)

enjoy and ask questions if anything's unclear

Thanks! Strength of component will be a valuable addition, if possible at some point.

Edit: Wow, this is really neat. You automated something that I've wanted for a long time. I think I might be the first to write a post on this for GSOM.

Better team results with Wall during the last two seasons in just 3 of 15 cases.

Perkins is the only guy Ibaka has better team results than.

Artest was the only guy Kobe had better team results than last season. 2010 he was way better than evertyone but Fisher. Quite a swing.

What's a good name for this? How about S4PM (same four +/-)?

Cool. I think if you color coded the offense and defense columns as well, that would make it easier to see at a glance the impact on offense vs defense.

Adjustment for SOS is coming tomorrow. I don't want to punish good players for being good (and having to play against opponent top players). It generally helps the star players like Kobe and LeBron, sometimes by as much as 5 net points. Color coding for offense and defense is also coming up.

Strength of schedule, according to opponent RAPM, is now factored in. And more colors. I also upped the possession requirement a little bit

Some of the players look pretty sick: Fields, Collison, Nowitzki, Foster, Ginobli, CP3

Kevin Love went from god-awful in '09 to almost all positive in the following years
Wade has quite a bit of red(=negative=bad) on defense over the last years. James Jones looks very good over the last 2 seasons. Luol Deng is all green on defense for the last 4(!) seasons.
Steve Nash never had a negative NET in 5 seasons....
KG in '08 and LeBron in '09 look like sick seasons: Better than everyone on both offense and defense

I haven't thought about a name, S4PM is fine with me. Thanks for the suggestion


R.I.P. Whitney Houston

Jerry, was Whitney Houston well known in Germany? Just curious.

This is other tool for looking at most valauble on a team, or lineup dependency.


Dirk has 9 comparisons +10 or better total for this season and last. Paul 7, Collison 3. Deng 3, Ginobili 6 (without the benefit of much time this season yet), James just 2, Nash 3, G Wallace 7, Milsap 5, Garnett 3.

Or you could total by minutes.

Andre Miller double digits better team results than with both his susbstitutes. Felton double digits better than one of his susbstitutes (Matthews), worse than the other (Crawford) so far.

Anderson went from -12 compared to Howard, to 0 to +20. Is this mostly random or mostly player adaption to role or coach adaption of player to role? Anderson is Howard's only negative comparison.

James +4 over Wade.

Carter double digits better than all his susbstitutes when playing with the same 4.

Barnes, the only player Artest / Peace has been had better results than this season or last.

Ed Davis dead last on RAPM estimate in the league and only better on S4PM than Kleiza. Do you wait for development or change usage or trade him?

Hickson all negative on his new team though mildly. Second worst on overall RAPM.

Glen Davis horrible fit in Magic's lineups, over 20 and 30 points worse team results than with his substitutes.

Batum with only one positive comparison over the 2 year period, a player since removed from the team.

K Leonard no positive comparisons. All but one double digit negative. Wrong draft pick or wrong usage or just give it time or balance these results with others?

McGee only one positive, to a lightly used young guy, and more big negatives than anygthing else. Do you change something or keep trying him and him in the same ways?

Patterson has moved from a negative comparison with Scola to a positive one. His minutes appear to be up this week but does it last? Does the performance last? Scola's minutes are bouncing around. Does that minute pattern continue or do they decide to go high or lower and on what basis- public stats or products of in-house techies? Would both paths point to the same conclusion or different ones?

Jerry did the hard technical work in developing his new tech tool. Will teams take notice and use it?

(Is this brief analysis worth something too? Don't you need both activities to get the value out of a tool? Is it considered self-evident? Is it assumed that everyone will get the same amount out of a tool? That there is no difference in how fast people find meaning or find / develop important or useful questions? It is not as easy or clearcut to cite this ability as say a specific advance programming language. But it is an important ability, I think. That can found be reading posts in a place like this. If a decisionmaker values productive use of data as much as production of data.)

Were all these teams already on top of these issues with their in-house technical products? If so, why are some of the problems this big and apparently this unresolved? You can't avoid all negtive performances of course, but you'd like to see more signs of recognition and mitigation with some of these big problems.

EvanZ wrote:Jerry, was Whitney Houston well known in Germany? Just curious.

Well, much more during the late 80's and during the 90's than recently, but still a big enough name to have the top article on basically all newspaper websites. She should be well known among people between 20 and 50 or so.

I was trying to think of a name rather than an acronym. Though, Evan's "same four plus-minus" flows off the tongue pretty well in a way that explains the stat. And, S4PM abbreviates that well in second references. I always liked how Bill James would try to name stats as what they were rather than using acronyms. Runs Created. Range Factor. Approximate Value. Disciples went more of a military route when they invented stats, turning everything into acronyms that made it seem like you needed to learn a new language to figure out what everyone was saying. Listeners shouldn't have to learn Klingon to understand why VORP was relevant.

Anyway, was playing around with ideas laying in bed last night. Was trying to do something with rotations, or with the concept of five because we're looking ultimately at five-man groups. If it becomes the consensus that JE's numbers here are very important in understanding basketball...maybe something like "quint-essentials" could be a term for them. Critical data that comes from seeing how different five-man rotations are playing out with four guys locked in and the other two compared head-to-head.

"That guy really hustles, and his quint-essentials are fantastic."
"Sure, that guys scores a lot, but his quint-essentials are surprisingly bad."

Like the double meaning...as the quintessential team player should grade out well in this stat, and we're looking at mutliple five-man rotations to make the assessments.

Just an idea. Same-Four-Plus-Minus really cuts to the chase well. Quint-essentials might get media run because the media likes putting puns in headlines...

Both names are good. Quintessential does have zing though it is not clear from the name what the specific comparison being made is, though it does suggest the underlying nature of the measure and its importance. "Substitute comparison" is bland. maybe too generic, but does provide more explanation of what is being measured.

EvanZ wrote:Jerry, was Whitney Houston well known in Germany? Just curious.

Her albums reached pretty much the same chart positions here as in the US, so I'd say yes

K Leonard no positive comparisons. All but one double digit negative. Wrong draft pick or wrong usage or just give it time or balance these results with others?

I think it's fine to be negative as a rookie, especially on the Spurs

Is this brief analysis worth something too? Don't you need both activities to get the value out of a tool?

yes and yes. Interpretation of the pure numbers is certainly necessary and useful. I just often don't have the time

Thanks.

Your technical skills are certainly yielding very worthwhile products. I suspect you could mine a lot from it if you had the time for that too, and perhaps even more incentive.

Some of the uglier player pages, in terms of color and thus efficiency, belong to
Ariza http://stats-for-the-nba.appspot.com/pm/245.html, with not a whole lot of NET positives since '09. It has to be mentioned that his deepest red comes from Chris Paul and David West
Raja Bell http://stats-for-the-nba.appspot.com/pm/7.html who was just horrible last year, with mostly double digit negatives
Hakim Warrick http://stats-for-the-nba.appspot.com/pm/218.html :A sea of red his entire career
Medium Baby Davis http://stats-for-the-nba.appspot.com/pm/675.html he had/has to backup some strong players. Still too much red for my taste
McGee http://stats-for-the-nba.appspot.com/pm/798.html with 3 NET positives in 4 years (one by 1 point, one by 2). Compared to Blatche he went -11, -12, -10, -15 over 4 years
and the maybe ugliest season of all time (err, since '08): Eddy Curry in '08 http://stats-for-the-nba.appspot.com/pm/235.html