Blog a Bull: An SB Nation Community

Navigation: Jump to content areas:


Sports blogs for fans, by fans.
Around SBN: Cal RB Jahvid Best Seriously Injured, Carted Off Field

Statistical Plus/Minus

Neil Paine from the Basketball-Reference Blog calculated statistical plus/minus for this season so far and a projected statistical plus/minus for this season based on the previous 3 seasons.  For more info about statistical plus/minus you should read Neil's blog post, but I'll try to give a brief description.  Statistical plus/minus was created by Dan Rosenbaum, who was a pioneer of adjusted plus/minus, in order to lower the noise of his adjusted plus/minus stats.  So you have fewer cases like Nocioni with terrible stats and a great adjust plus/minus.  Rosenbaum determined weights for the various boxscores stats based on how well they predicted adjusted plus/minus.  So what you have is a boxscore metric like PER, Win Shares, or NBA Efficiency, except it's trying to predict point differential.  In overly simple terms statistical plus/minus is the adjusted plus/minus a player should have based just on their boxscore stats.  It's not meant to be a player rating system by itself, but I think statistical plus/minus is very useful to compare to this season's adjusted plus/minus,and to compare to a stat like PER.

Player          Tm      Pos     G       Min     SPM     projectedSPM
rose,derrick chi PG 54 1989 -1.71 -2.16
gordon,ben chi SG 54 1944 0.15 0.57
deng,luol chi SF 43 1484 -0.66 0.83
thomas,tyrus chi PF 51 1307 1.22 0.38
nocioni,andres chi SF 53 1281 -1.88 -0.74
noah,joakim chi C 52 1115 3.66 1.15
gooden,drew chi PF 31 920 -2.81 -2.67
hughes,larry chi SG 30 792 0.21 -0.95
sefolosha,thabo chi SG 43 735 -0.44 -1.68
gray,aaron chi C 45 662 -0.83 -0.58
hinrich,kirk chi PG 23 611 1.34 1.77
CHICAGO -1.70 -1.99
salmons,john sac SG 53 1980 0.09 -1.04
miller,brad sac C 43 1358 1.39 1.62
thomas,tim nyk SF 36 771 -0.21 -0.33

Star-divide

Few of the statistical plus/minus scores surprised me.  Rose's strong negative doesn't surprise me.  He's just been statistically poor in so many areas of the game.  The difference between Tyrus, Miller, and Gooden, despite similar PERs is very expected.  Noah's great score is expected and fits his adjusted plus/minus scores over the last two years.  I was happy to see that Salmons had a positive number this season.  That at least suggests that there's nothing in his stats this year that would lead to his very negative adjusted plus/minus.  Nocioni's very negative statistical number reinforce that his positive adjusted plus/minus this season was a fluke.

The one real surprise is Hinrich's numbers.  I don't fully understand why his score is that good, yet.  But, I can't just dismiss it because Hinrich's adjusted plus/minus is also positive this year.  And I've seen more times, unlike last game, where the team has genuinely played better with him on the floor.  I can't dismiss the notion that the team plays better with Hinrich as easily as I could before. 

Take a look and read the blog post, but remember that statistical isn't intended to be a rating system by itself.  It's a piece of the puzzle that can help you interpret other stats.

FanPosts are user-created posts from the BlogABull community, and are to be treated as the opinions and views of that particular user, not that of the blogger or blog community as a whole.

8 recs  |  Comment 61 comments

Story-email Email Printer Print

Comments

Display:

Thanks Scotter. Neil posted this information on the APBR board before

he posted it on B-R.com blog, I almost posted it, but I’m glad I didn’t because Neil more fully explains things in the blog post and your analysis is generally more concise and easily understandable than my own. Great stuff.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 23, 2009 9:13 PM CST reply actions   0 recs

I almost posted it too, but didn't for similar reasons.

Looking back on those numbers. Gordon was the most positive offensive player at a little +2. Hinriich and Rose were the only other clearly positive offensive players at +1. Thomas and Noah were the only significant defensive players, both were +3 on defense.

by Scotter on Feb 24, 2009 12:53 AM CST up reply actions   0 recs

that seems to make some sense based on my own subjective evaluations of things from

watching things. I’m just surprised how invisible Deng is in these numbers. He seems neither good nor bad, but he’s always been great in adjusted plus-minus, is he a little like Battier in that he does little things that wouldn’t show up as much in statistics so statistical plus-minus would underrate his contributions? I noticed he deflects a lot of passes that lead to steals for his teammates and generally seems to make life uncomfortable for opposing wings. He’s not a dominant defender, but just seems like a guy who does a lot of the little things that go unnoticed.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 24, 2009 1:02 AM CST up reply actions   0 recs

I would suspect part of Deng's adjusted plus/minus is two things.

I think he means more to the Bulls offense than he would to other teams. The Bulls offense is more dependent on Deng’s points in the paint than most teams would be. That Deng gets most of his points off the ball also wouldn’t show up in statistical plus/minus, and that might separate him from a 1 on 1 scorer with similar scoring stats that doesn’t have to be guarded without the ball.

The most significant aspect of Deng’s defense is forcing his man to take more difficult shots (Somewhat like Battier), which would show up in adjusted, but not statistical plus/minus very well.

by Scotter on Feb 24, 2009 1:19 AM CST up reply actions   0 recs

That sounds about right.

Do you think, then, that Deng would be less of an adjusted plus-minus contributor if placed on another team that had scoring in the paint from a big man? Do you think this had anything to do with his struggles at the beginning of the year because they were clearing the paint for Rose to penetrate and not really having Deng moving and cutting into the lane?

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 24, 2009 1:29 AM CST up reply actions   0 recs

That's what I immediately began questioning.

Would Deng be as positive on another team? I’m just not sure right now. Based on the offense/defense adjusted splits from last season, most of Deng’s +/- is offensive. So I do question whether Deng would be as positive in an another setting, but I’m not sure which side to come down on.

I think a decent portion of Deng’s struggles are offensive design problems. He was getting significantly fewer shots in the paint. He also shoots jumpshots better on the move, rather than spot up. Deng really is pretty much the anti-Nocioni. I think the offense affected other players as well because the only players succeeding early on were the guys that can score 1 on 1. That’s part of Deng’s 4th quarter problems as well. For the 1st 3 quarters, and especially in the 1st half, teams focus on running their offense, but come crunch time teams start to devolve into more 1 on 1 play. I also blame Deng playing for Britain to a degree because he’s playing differently for them than he does in the NBA, and I don’t think he makes the adjustment well because it seemed to affect him last season as well.

by Scotter on Feb 24, 2009 2:15 AM CST up reply actions   0 recs

He's already imho better than

Battier, better because he’s a much better offensive player, but Shane’s still the superior defender….hopefully Deng can grow there to be Battier like on D, because if he does that, we’ll all be very happy (Deng’s not a bad defender, but has some room to improve).

by majoyenrac on Feb 24, 2009 8:10 AM CST up reply actions   0 recs

We'd definitely agree on that.

I’m convinced that Deng just needs the information and coaching that Battier gets in this regard. Well, the whole team does, anyway.

Viva la nuance! Reading comprehension rules!!!

by tyger1147 on Feb 24, 2009 8:12 AM CST up reply actions   0 recs

Scotter, considering that LBJ is around 15

do the Bulls have anybody worth keeping (besides Rose assuming he’s worth keeping)?

by hlac on Feb 23, 2009 9:16 PM CST reply actions   0 recs

I think the key guys to keep are Tyrus 1st and Noah 2nd.

Their impact on defense is significant even with all there flaws. They should get even better on that end. Tyrus’s offense will continue to get better. Noah obviously has a much lower offensive ceiling, but he can get better on that end as well. Those are the guys who can be significant difference makes.

by Scotter on Feb 23, 2009 11:24 PM CST up reply actions   0 recs

I would be happy with Noah

if he could get to +75% FT, as he will get fouled a lot when going for rebounds and put-backs.

Maybe we can?

by Granny Waiters on Feb 24, 2009 1:27 AM CST up reply actions   0 recs

Seriously

While others have said it, when are you going to work for the Bulls either in scouting or coaching….great job! great analysis.

by majoyenrac on Feb 24, 2009 8:12 AM CST up reply actions   0 recs

And to think two years ago

the core of Hinrich, Gordon, Deng, Wallace was supposed to give us 50 win season after 50 win season.

by YaoPau on Feb 24, 2009 2:28 PM CST up reply actions   0 recs

Scotter, your next assignment (should you choose to accept it)

develop something similar for coaches. We need this fast. By this weekend would be fine.

by hlac on Feb 23, 2009 9:23 PM CST reply actions   0 recs

Pretty reasonable

Gerald Wallace as the 12th highest in adjusted +/- is kinda odd, no?

Vinny Del Negro interviewed for the job today. I mean come on! Nobody else thinks this is nuts?
by Juiceboxjerry on Jun 6, 2008 4:21 PM CDT actions actions 0 recs

by Ozzie Montana on Feb 23, 2009 9:24 PM CST reply actions   0 recs

statistical +/-*

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 23, 2009 9:25 PM CST up reply actions   0 recs

Also, the thing that really jumps out is that Noah is our best player by a pretty wide margin

and last year he was 2nd in adjusted +/- behind Deng and he’s got the best ORtg on the team, of the players that have been on the team all year (somewhat mitigated by his low Usg%) and the second best DRtg on the team (excluding the new guys). It just goes to show how under-appreciated Joakim is amongst many, many fans.

Finally, I’m a bit surprised by how low Deng’s SPM and pSPM came in because he’s traditionally done great by adjusted plus-minus every year. He’s doing very well by that metric again this year according to basketballvalue.com, again best on the team this year.

Finally going to Sports2’s points earlier about the trades for Salmons and Miller… Salmons has been a positive SPM this year (.09) and Miller (a consistently dominant adjusted +/- performer) is a positive SPM performer and pSPM scorer. We sent out almost all negative SPM performers (Thabo was slightly negative, Noc and Gooden were very bad and we sent them out. Hughes was not negative by SPM, but was negative by pSPM and has been hugely negative by adjusted plus-minus and anyone who has a pair of eyes and watched this team play could see that Larry, Andres, and Drew were all affirmatively harming the team when they played.) I don’t think that Salmons is an adequate replacement for BG, but I am heartened by Kirk’s solid performance in adjusted plus-minus, stat plus-minus, and the Bulls HAVE looked better overall as a team since he’s returned. This makes me think that maybe a platoon of Kirk, Salmons, and Rose can work in BG’s absence. It’s not my ideal solution, but if it leaves us more flexible for 2010 to pursue a real star to pair with Derrick, Deng, Tyrus, and Noah then it’s a sacrifice I’m okay with making.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 23, 2009 9:55 PM CST reply actions   0 recs

I don't think Salmons is an adequate replacement for Gordon either.

But, I’ve come to suspect that the difference isn’t worth the costs required to keep him. The team can survive with Rose, Salmons, Hinrich for a year, that’s a lot different than Rose, Hinrich, and Thabo. If Salmons starts and plays well then he probably opts out of his contract, which clears more cap space.

by Scotter on Feb 23, 2009 11:59 PM CST up reply actions   0 recs

right, I've been swayed to this way of thinking the more I've thought about it

I guess you could call it me going through the “bargaining”, “depression”, and “acceptance” stages of grief in the loss of Ben Gordon as my starting two guard. Sigh.

Pax better land a big fish in 2010 though. Otherwise, I will be PISSED.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 24, 2009 12:11 AM CST up reply actions   0 recs

Thanks Scotter...

…I’ve been wary of +/- (particularly when based on a single game), but this looks like it can be helpful. I always find it more fun when metrics tell us something we wouldn’t expect just from the eyeball test, especially when it contradicts something people vigorously claim.

Particularly like the info on Hinrich, since the negative groupthink on this board seems to have set its sights on him now…

by T Maple on Feb 23, 2009 10:13 PM CST reply actions   0 recs

It all has to do with Rose.

Hinrich doesn’t play well with him. Either Kirk is on the floor as a PG when Rose is on the bench or he is controlling the ball and Rose is standing around the three point line out of the action. It is hard to get excited about someone who looks the best when our future star isn’t involved in the game. Add in his minutes per game and contract and you get a whole lot of resentment from anyone with the future in mind.

Everything I post is speculation. I have no insider information nor ideas deemed concrete enough by those who are self-elected to regulate post content.

by cranscape on Feb 23, 2009 11:04 PM CST up reply actions   0 recs

Maybe Rose needs

to improve his game so that he is still useful when playing with Kirk. If it’s true that Rose is just hanging around the 3 point line, maybe he should either work on his 3 point shot, or get more involved some other way.

by runningman on Feb 23, 2009 11:30 PM CST up reply actions   0 recs

you don't make Derrick Rose adjust his game to Kirk freakin' Hinrich. Hinrich needs to

adjust to Derrick “franchise player” Rose.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 24, 2009 12:15 AM CST up reply actions   0 recs

I don't think it's true that they don't play well together.

There’s certainly know statistical evidence that the team suffers when they play together. It’s certainly not in the player pairs data. I’ve come to the conclusion that we should give up on Derrick being a Chris Paul or Nash, handling the ball 90% of the time. I don’t have significant problem with Derrick playing off the ball anymore. He actually moves without the ball pretty well, something I noticed een watching his playa t Memphis. And lately he has had more opportunities to get to the rim by catching the ball on the move off of cuts. I’d still never want them starting together, but if they’re paired together for 10-15 minutes a game it’s not a big deal.

by Scotter on Feb 23, 2009 11:39 PM CST up reply actions   0 recs

82games ranks every player on the roster

from most point guardish to most centerish in order to track those stats.. Roland said they have more accurate stuff that they sell to NBA teams, but for the public stats their position rankings are the best they do. Rose got the most point guardish destination on the Bulls. The counterpart data isn’t good enough to track when Hinrich is the point guard offense, but guarding the other team’s SG.

by Scotter on Feb 24, 2009 12:30 AM CST up reply actions   0 recs

Right

I dont know how indepth they go, but a computer could guess at that. I’d imagine not that deep.

However, that doesnt have anything to do with the difference in his PER by position, much better as a SG. And also, even if it is counting the wrong counterpart when he is the PG often times that person is going to be guarded by Derrick or Ben(“bad” defenders).

You cant get carried away drawing conclusions from it, but both his & his counterpart PER is probably somewhat tied to (until recently) playing PG mainly 2nd unit on 2nd unit, and Kirk actually being able to play some pick & roll D. The difference in eFG is odd, coincidence? However the SG counterpart PER is pretty damn high(3pts higher then BG’s) and definitely an indictment of what the team is doing when he is out there with Derrick.

by reprisal on Feb 24, 2009 1:08 AM CST up reply actions   0 recs

If you look at the Player Pairs data.

Hinrich has played a nearly identical number of minutes with Rose (321) and Gordon (331). The team’s point differential is 103 to 100 with Hinrich/Rose, and 101 to 92 with Hinrich/Gordon.

Something else to consider is that Gordon and Rose’s counterpart PER are artificially low because of the combination of them allowing dribble penetration combined with Gooden and Nocioni’s porous defense. Opposing guards had low PERs because they were busy feeding their bigs for easy buckets for the 1st two months of the season. As the Bulls interior defense has gotten better, Gordon and Rose’s counterpart PERs have steadily risen.

by Scotter on Feb 24, 2009 1:42 AM CST up reply actions   0 recs

I had thought of that awhile ago.

Didn’t really make the connection to Kirk only playing since mid Jan tho. That certainly could explain some a decent amount of it. 17.5 is pretty bad regardless tho.

And Rose/Gordon is 101/104…sigh Derrick really is awful on D.

by reprisal on Feb 24, 2009 1:56 AM CST up reply actions   0 recs

I would run alley-oop plays with Rose on the receiving end,

just because it is so cool to watch a point guard finish that way. :)

Maybe we can?

by Granny Waiters on Feb 24, 2009 1:31 AM CST up reply actions   0 recs

Scotter thuogh

Have you run early Nash years, D-Will’s first year, etc etc etc. Rose is 20 yrs old and seems his biggest flaw right now is aggressiveness, followed by defense…..both are fixable (aggresiveness will come naturally as Rose improves/matures) and defense is all about effort and time….he’ll be there (too athletic to not be in my book)…..

I think it’s too early to state what 20 yr old Derrick can and can’t do……or will or won’t be.

I do think though and have thought that I don’t mind Derrick at the off guard spots at times….he seems adept at that role and it forces him to being a bit more aggressive at taking the ball to the hole than when he’s the PG for the entire game (it also seems if he plays some 2 then reverts back to the 1, he’s more aggressive at the point as well).

by majoyenrac on Feb 24, 2009 8:18 AM CST up reply actions   0 recs

i don't particularly like rose at the off guard

but what i hate most about it is hinrich ignores him, much like we saw what he did to bg in the 3rd quarter on sunday.

"They should. They better. I'm Vinny Del Negro!"

by Jaina on Feb 24, 2009 8:25 AM CST up reply actions   0 recs

Yeah

But I don’t know if that’s a normal Kirk move (ice somebody out). I think Kirk was trying to get something going because he was awful and just added fuel to the fire by not getting himself on track and icing out BG, everyone else after he continued to suck.

That’s more a VDN issue though, after Kirk bombed in the first half and continued to bomb in the 2nd half, VDN should have pulled him, but didn’t (and I am a Kirk fan).

by majoyenrac on Feb 24, 2009 8:50 AM CST up reply actions   0 recs

Thanks for the heads up on the new stat, but I'm skeptical about this one

The W/L predictions were shockingly accurate… until he explained he used the exact minute breakdown of each team. If you give any schmo who follows basketball the exact minute breakdown of each team for the next season, he could beat any computer projection too.

I don’t know everything about the more advanced stats that have come out in past years, but it seems to me that any of these One Number To Explain Player Value stats are a waste of time. Just an amateur guess here, but I think basketball stats will have to eventually move to quantifying what unique skills a player brings to a team, and then figuring out how the pieces go together, if they’re ever going to mean anything. For example, I could start a business and hire ten of the smartest managerial talents in the city, but if we don’t have someone who knows accounting, or can run machines, or can engineer new products, we don’t have much, do we?

by YaoPau on Feb 24, 2009 9:13 AM CST reply actions   1 recs

Was this not clear enough?
It’s not meant to be a player rating system by itself

We have every right to dream heroic dreams. Those who say that we're in a time when there are no heroes, they just don't know where to look.
Ronald Reagan

by snley on Feb 24, 2009 9:59 AM CST up reply actions   0 recs

Well then what's the point of it?

Clearly Rosenbaum was trying to combine +/- and stats into one in an attempt to create a better standalone statistic. If that wasn’t his aim, then he would’ve left those stats separate like a lot of us do when we list PER and TS% or adjusted +/- together.

It’s cool to have a new advanced stat, but it’s worthless if nobody knows what it means.

statistical plus/minus is the adjusted plus/minus a player should have based just on their boxscore stats.

Simplified description or not, we all know that leaves a lot out.

by YaoPau on Feb 24, 2009 12:12 PM CST up reply actions   0 recs

no one stat is going to be able to cover everything.

some are more comprehensive than others. it may be a “better” standalone statistic than others, but it still needs to be taken in context with other things.

"They should. They better. I'm Vinny Del Negro!"

by Jaina on Feb 24, 2009 12:38 PM CST up reply actions   0 recs

Well, what you really want is to create a blind test, not a retrospective data fit

Like, if you develop the statistic, then plug it in to actual minute distributions and find that you’re right on the money, you’ve got something very good.

If you develop the statistic by trying to fit historic data, you’re not necessisarily going to have future success.

——————————-

In any case, I’m mildly peeved because Dan and Neil actually thought of and did this while I merely was sitting here thinking about it the last couple of weeks. I really was… I’d downloaded the data from basketballvalue and scrounged up my Stata CD and everything. Hats off to them for doing something cool why I fart around.

by Sports2 on Feb 24, 2009 10:00 AM CST up reply actions   0 recs

He does suggest

that it is best used to complement adjusted +/- . He also suggests somewhere that he feels it shows some promise as a defensive metric, of which there is certainly still plenty of room for improvement.

But ya, raw box score data is pretty limited to project what you describe. Stuff like game play by play logs and shot charts that can begin to offer more insight are nearly as readily available in a digestible format and would require more parsing to get the data into a useful format. And those things don’t even provide that much more detail.

Maybe teams internally have the resources to collect better raw data to pull from? (e.g. Mark Cuban’s Jason Kidd stat, lol)

by reprisal on Feb 24, 2009 12:54 PM CST up reply actions   0 recs

Thanks for all the effort put into gathering these statistical +/- stats. They provide valuable performance indicators, however

When looking at projected numbers, I think they don’t have as much predictive value when looking at our younger players, who are just getting opportunities to get consistent minutes for the first time. I don’t have the stats, but my eyes have shown me that TT and Noah have been playing much better and their arrow should be pointed upward the more they gain experience. The same goes for Rose. Using these measures to determine how veterans may perform has validity because the likelihood of some vast change in either direction is minimal.

It also reveals what most of us already knew by observation, that Gooden and Noce were disasters, especially defensively and the young bigs improved our defense when they were on the court in spite of making mistakes on covering the pick and roll and adjusting on the fly. I was not surprised that Brad Miller and Kirk had positive stats, since they both have veteran skills which don’t always show up in the boxscores. Granted Hinrich had a horrendous game Sunday (although his +/- was 0), but overall he does provide solid defense against most opponents with his ability to stay in front of them unless they are much quicker then him.

Miller, in spite of his Grayness speedwise, has 3 big S’s going for him. He’s STRONG and can hold position in the paint which helps him get rebounds over more athletic centers, he can SCORE from anywhere on the floor and he has lots of basketball SMARTS, demonstrating many of the tricks that P J Brown used to provide. His court awareness and passing skills will greatly aid our younger players and should create more open looks for Gordon.

I love Deng, but these numbers may lead one to question if he wasn’t vastly overpaid for the skill set he brings. Maybe the new additions will help his game especially with Miller’s ability to play the high post, thus opening up more space for slashers like Luol.

Gordon is who he is, an amazing scorer and a below average defender. But no one can honestly make a case that he isn’t our most effective and explosive offensive weapon. Without him, our ability to put points on the board will greatly diminish and thus JR needs to come up with enough sheckles to keep him from jumping to greener pastures.

Once again Scotter has knocked one out the park with the bases loaded. Please, Vinny put him in charge of your rotations, especially with the game on the line. Statistically, I would guess the improvement in results would be something like 99.9%

If you can't answer a man's arguments, all is not lost. You can still call him vile names.
Elbert Hubbard

by Tyrusmancrush on Feb 24, 2009 1:22 PM CST reply actions   0 recs

The value of the projected numbers

I agree it’s more or less useless to make a uniform projection for rookies. I’d guess Neil Paine was just doing that to have some placeholder data so he could make team-based projections.

So I wouldn’t look at that at all.

The value in the projection vs. actual production for more veteran players is pretty interesting though. If I follow, what’s being done with the pSPM is to use past data to predict the actual SPM a player would have in 2008. What I find interesting about it is that it says most of the Bulls are playing pretty close to how they this would predict them to. Looking at Deng, for example, it suggests he’s performing below past-performance based, expectations, but not vastly below them. He’s underperforming less than, say, Tyson Chandler or Jamal Crawford.

Of course, he’s still underperforming and his predicted number was somewhat surprisingly low to start with

He’s not

by Sports2 on Feb 24, 2009 2:15 PM CST up reply actions   0 recs

You know... after meditating on this a bit more... not sure I like it much.

Someone correct me if I’m wrong, but…

So what you have is a boxscore metric like PER, Win Shares, or NBA Efficiency, except it’s trying to predict point differential.

See… I think the logic for a box score metric breaks down pretty severely when you cross the threshold into teammate and opponent adjusted point differential methods.

That is, what’s the underlying problem we’re trying to solve?

We’re trying to properly account for a player’s value toward winning a game. And the game is determined by point differential

The logic behind box score weighting schemes is that we know box score stats affect the game, but they don’t directly translate to the direct determinant, which is point differential. Weighting things like rebounds, TOs, assists, shooting percentage, etc. is only valuable insofar as they contribute to team point differential.

To me, this ultimately suggests that a well-constructed SPM must be less accurate than a well-constructed APM.

  • In the case of SPM, you’re making an estimate of a every player’s contribution accounting for every other player’s contribution, but measured in terms of an estimated point differential based on linear weights.
  • In the case of APM, you’re making an estimate of a every player’s contribution accounting for every other player’s contribution, but measured in terms of… the actual point differential.

APM is based on actual observations of the determinant variable (point differential). SPM is based on synthetic observations of the determinant variable. Because of that, my guess would be that, given enough data, APM is going to give a more accurate result.

by Sports2 on Feb 26, 2009 6:18 AM CST reply actions   0 recs

I think you're right Sports2. I'm not sure that SPM has much value outside of reconstructing

seasons where they don’t have the data to run the lineup numbers for APM. It’s fun to look at some of the numbers though and see where they lineup with APM and where some guys over and under achieve relative to their APM.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 26, 2009 8:12 AM CST up reply actions   0 recs

But, the difference between SPM and other linear weights

systems is that it’s not intended to be a player rating system. It was intended from it’s creation to be used in conjunction with adjusted plus/minus. The problem with adjusted plus/minus is that it can be very difficult to conclusively figure out why a player has the APM that they do in many cases. With the amount of noise in APM that’s a problem. SPM is there to either reduce the noise by being combined with APM like Rosenbaum does. Or it’s there to compare against APM. SPM is there to help you figure out how accurate an APM score is. It’s primary value is in interpreting other stats, and I believe it to be very valuable for that.

SPM is also different from most linear weights systems because it isn’t logic based. SPM weights were the product of a regression of APM. The regression found correlations like players shooting more free throws tend to play better defense. Players that turn the ball over tend to be less effective defenders. Three points shots have offensive value beyond the value of the points scored. No logic based linear weights system would have such diverging numbers for Gooden and Tyrus. Every stat has an offensive and defensive value. SPM is very useful as long as you’re not treating it like PER. As Oliver argues we need to get away from player ratings system anyway because the majority of players are heavily effected by role and team. PER can give you the top 5 players in the game. It doesn’t help you figure out if you would rather have Gooden or Tyrus on your team.

by Scotter on Feb 26, 2009 1:22 PM CST up reply actions   0 recs

Respectfully... it's being used as a player rating system, and damn the intent.

It’s not intended to be a player rating system, which you do clearly acknowledge in your original post… right before you employ it as a player rating system :) and cite Neil’s post that gives a pretty clear impression, despite the caveats, of using it as a player rating system and predictor. Not simply a complement to APM. :)

I mean, I understand that you were careful in your post to cite your understanding of the stats in context of your of the lastest APM stats, but the quick read of your post is pretty misleading…. despite the caveats it’s really presents like a list of “who’s good” based on the latest and greatest stat.

And yeah, I understand that the SPM weights are the product of a regression of APM, but one has to be very careful when acknowledging he doesn’t understand the logic behind the correlations being employed. The data sample is relatively small (only a couple years), the player cohort and number of variables is so small, and the game itself is changing in important ways (ask Kirk Hinrich’s defense about how things are called vs. a couple years ago!) that I’d expect the weights Dan originally found to be fairly unstable. I could, of course, be wrong about this, but I’d like to be proven wrong :)

Let me put it a little bit differently. I don’t see how one can say it’s not intended to evaluate a players’ contribution, but then use it as a check against a measure of a player’s contribution. It either has something to say about it or it does not.

I think the reality is that it does. Operate as a player rating tool So lets just admit that it does, and try to understand the hows and whys, rather than say it’s not and then use it as one anyway.

by Sports2 on Feb 26, 2009 2:47 PM CST up reply actions   0 recs

Realistically this is devolving into the offensive rebounding debate all over again.

Neil is exploring what statistical plus/minus can do, which is good because there hasn’t been much work done with it. It was an exploration of a stat that very few people knew about unless they read Rosenbaum’s 82games article. I think you’re imposing your view point on both what Niel and I wrote.

Dan has clearly hinted that they’ve made significant advances in both APM and SPM and much like Dean Oliver was holding stuff back that he didn’t publish even before he got the NBA job, unfortunately that work isn’t going to be made public for a long time if ever.

The stat is intended to identify the difference in APM a player with those boxscore stats would typically make. That doesn’t make it a player rating system anymore than ORtg and DRtg are player rating systems. That’s what is, and I didn’t see anywhere in my original post where I used it for anything, but that. There’s a difference between evaluating a player’s contribution and a player rating system. That might be semantics to you, but the difference in connotation and mindset is important.

by Scotter on Feb 26, 2009 3:56 PM CST up reply actions   0 recs

It's not semantics to me, which is why the difference in "intended" and actual use isn't lost on me.

Look, I’m not trying to rain on your parade, but whatever you want to call it, Neil was taking SPM and using it to rank out teams and players, and testing it out as a predictive metric. I don’t see anyone doing that with ORtg or DRtg. I don’t have a problem at all with that. That’s how we learn. But c’mon… Neil’s post isn’t comparing SPM and APM at all. Your post does, but the first thing one sees is a big table listing out SPM for every player. So yeah, connotation and mindset is important, I agree.

by Sports2 on Feb 26, 2009 7:24 PM CST up reply actions   0 recs

Neil did the the same thing with

ORtg and DRtg before to try and predict win totals on the APBR board. He’s very interested in whether stats can make accurate predictions. I’ve read my post again a couple of times now, and once again I’d say your going out of your way to impose your view point on what I posted.

by Scotter on Feb 26, 2009 8:09 PM CST up reply actions   0 recs

Once again I'm not trying to impose my views on anyone

That’s a pretty silly thing to say in a blog comment thread, by the way.

by Sports2 on Feb 27, 2009 7:32 AM CST up reply actions   0 recs

OK, I'm not "imposing my view point on what he posted"

Is that pedagogically correct while still tortured into absurdity enough for you? There’s no imposition to be had by two people somewhat disagreeing (much less than one of them appears to think) in a set of freaking comments on a freaking blog.

by Sports2 on Feb 27, 2009 12:08 PM CST up reply actions   0 recs

wow?

Viva la nuance! Reading comprehension rules!!!

by tyger1147 on Feb 27, 2009 9:07 PM CST up reply actions   0 recs

I think it's being as a descriptive stat, not necessarily as a player rating system.

It can be used by people as a player rating system if that’s the spin they choose to put on it, but it’s no more a player rating stat than TS% is a player rating stat or even FG%. It’s a fuzzy description of the net effect of what happens on the court when the player is on it. If people want to use it as a player rating system, they can, but that doesn’t make it a player rating statistic in the abstract.

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 26, 2009 4:56 PM CST up reply actions   0 recs

d'oh!

being used as a descriptive stat*

Vinny discovers frontcourt by accident. Someone re-smash Gooden’s groin!
- your friendly BullsBlogger

by fundamentallysound on Feb 26, 2009 4:56 PM CST up reply actions   0 recs

To Scotter, Sports 2, and fundamentallysound

I would say that one of the more helpful uses of player statistics would be for the scout and player evaluator to use in watching and evaluating players. The statistics would help the evaluator organize what he observed into a thoughtful assessment of the player.

Do you agree? (or does this go without saying)

by hlac on Feb 26, 2009 11:10 PM CST reply actions   0 recs

It goes without saying... but...

the trick is how to use them. We all know our eyes can be deceiving, and we all know stats can be deceiving. At the same time, we know they can be useful. So how do we organize our observations when different stats tell us different things, and potentially, our eyes tell us different things too?

With stats especially, we face a little bit of difficulty because numbers, especially poorly understood and complexly derived ones, often give an appearance of certainty when they don’t exist (ht: to David Li’s – not the Knick’s David Lee!- copula function that bewitched and managed to nearly single-handedly fuck the financial system).

I don’t think anyone actually disagrees on this point. Scotter seemed to take offense because, I guess, he felt he’d properly qualified his original post. I don’t think his qualifications quite got the point across to a broad audience. Simply because a broad audience isn’t going to get that point. (It has nothing to do with the intelligence of the audience, by the way). Very much in the same way that if I unleash a new puppy in my house, my kids would go nuts playing with it despite sober addresses delivered about proper use, responsibility and care of the puppy.

by Sports2 on Feb 27, 2009 7:59 AM CST up reply actions   0 recs

My two cents

I think what the purpose of a post like this was to point fans to a new system of interpreting a widely used but imperfect stat that’s only value is to rate players (to said fans). The debate judging its merits was informative. I’m very much on board with you re APM’s significance. But as a stand alone stat, it does seem to have a pretty good “laugh factor”. If SPM is mostly good at reducing that factor, hooray, that is big news to someone like me. I’m not trying to earn a living from my posts or pursue a career in basketball. What I am interested in is if my sentiments about the value of, say, Joakim Noah are borne out by the stats.
     In the end, SPM is probably not influencing VDN’s or Pax’s decision making. We all have biases, and much like our religious or political biases, we will tend to discount information that disputes those biases, and place a premium on ones that support them. So, most us are looking at these systems and judging them by how well their ratings line up with our intuitive ratings. In the end, all they’re really capable of is giving us an appreciation of other players around the league who rate highly in a system that seems to reflect our thoughts on how to win basketball games.
    So if a system says JR Smith is batter than Battier, I’m inclined to reject it. If it rates a Billups higher than Iverson, I’m intrigued, and I’m immediately looking to see how it rates Bulls’ players. If it largely “ranks” them to my liking, you may or may not see me use that stat to support a post here. Better that than “Luol Deng sucks the big one”.

by California Al on Feb 28, 2009 9:19 AM CST up reply actions   0 recs

Comments For This Post Are Closed


User Tools

Welcome to the SB Nation blog about Chicago Bulls.
Start posting about the Bulls »

Join SB Nation and dive into communities focused on all your favorite teams.

FanPosts

Community blog posts and discussion.

Recommended FanPosts

Small
The Bulls and Young Talent
Blogabull_s_small
Sign up for BlogABull game previews

Recent FanPosts

Sports
I'm Happy to Have a Break From Tyrus. He's gone anyways.
Small
Game Preview # 6: Bulls vs Charlotte Bobcats
Small
Why isn't there more talk of LeBron...
Small
Contract/Cap Question for BAB
Jayhawk_small
Game Preview # 5: Bulls at Cleveland Cavaliers
Small
We've Got To Rescue D. Wade
Blogabull_s_small
Chad Ford Rounds up extensions and team options for recent draftees
Pippenmug1_small
Game Preview #4: Chicago Bulls vs Milwaukee Bucks
Drose2_small
Start of trade talk....

+ New FanPost All FanPosts >

SPONSORS


Guy who does everything

Blogabull_s_small your friendly BullsBlogger