Plus-Minus

by John Raffo on Oct 15, 2009 9:23 AM PDT

Maybe it's just me but is it a new thing to list player's plus minus in the boxscore? Both NBA.com and ESPN list the stat (which may or may not be from the same source), the LA Times and SBNation do not. Perhaps I haven't been paying attention, but I only noticed this phenomenon last night when the game was not televised and I couldn't get the radio broadcast, but I could scan the quarterly stats...

But I wonder at it's significance? If you look at the boxscore from last nights' game (here) you'll see some interesting details: While, judging by most reports, Al Thornton had a nice game, including eleven points in a row in the fourth quarter (8 for 11, 4 rebs, 18 points in 24 mins) his plus-minus was a sad -6. Ricky Davis on the other hand who also played 24 minutes, was 2-7 with 7 points and was a whopping +9. Craig Smith had only two points, fouled out and was +10. On the other side of the floor Theo Ratliff played 8 minutes didn't score, had 1 rebound and was also +10!

Is the plus-minus meaningful at all?

Tweet Comment 24 comments | 0 recs |

Do you like this story?

Comments

A few thoughts

NBA.com has had plus/minus for a few years. ESPN added it recently – I’m not exactly sure when.

There are many issues with plus/minus as such. The Al/Rasual situation last night being a prime example.

Al was 8 for 11, 18 points and 4 rebounds in 24 minutes. Sool was 1 for 7, 3 points and 1 rebound in 25 minutes. From those numbers you’d conclude that Al had the much better game, but Al was minus 6, and Sool was +4 (bearing in mind that most but not all of the time they were not on the floor together).

Now, it goes without saying that points and rebounds aren’t the sum total of a player’s impact on the court. It’s entirely possible that Butler’s defense and overall game compensated for his misses, or that poor defense from Al completely negated any scoring positives. I didn’t watch the game, didn’t listen to the game, so I don’t really know.

I do know that Butler played most of his minutes during the first and third quarters, while the Clippers were playing well as a team. And I know that there are five players on the floor at a time, not just one. And this is the single biggest problem with pus/minus. It seems pretty clear that last night, Butler’s positive rating had more to do with his teammates (he was out there with Baron and Blake Griffin and DeAndre Jordan) than it did with him. At the same time, Al was playing with Mardy Collins and Brian Skinner and Craig Smith. So while the Clippers were minus 4 with Al on the floor, I think it’s safe to say that they would have been minus a lot more than 4 had he not been there. It looks to me like he was carrying that unit, keeping them in the game.

A few stat geeks have developed a thing called adjusted plus/minus, which tries to compensate for this particular problem. Basically, it looks at who else is on the floor with you, what their plus/minus is, and to isolate the individual in the numbers. It’s a good idea.

But yeah, pure plus/minus has issues. It is useful for tracking the truly impactful players (the superstars who have the most influence on the game). But it can be misused, especially when comparing players who are playing with substantially different units.

In this world, you must be oh so smart or oh so pleasant. Well, for years I was smart. I recommend pleasant. - Elwood P. Dowd

by Steve Perrin on Oct 15, 2009 9:55 AM PDT reply actions

Yeah...

I was hoping to drag up some opinions from some of the resident stat-heads… especially since my initial thinking was that plus-minus has no place in a box-score. It begs more questions than answers.
But it’s an interesting number. I would be curious to look at career plus-minus from big time NBA stars present and past. Is there a source?

by John Raffo on Oct 15, 2009 11:52 AM PDT up reply actions

i always find it interesting myself

not form the individual player perspective, but from the unit by unit and player vs player plus minus.

this can give a coach and idea of how to assemble the units, which units work best together and how to weight each unit’s minutes.

also when two players match up for a game like baron and monta ellis the other night. you can get real idea of who dominated the matchup. it felt really nice to read baron +16/monta -13, i know its only preseason, but to dominate a little pg who can score like that, might be a good indication that baron is on the right track.

No gang...... Independent....

by Clipochistic on Oct 15, 2009 10:24 AM PDT reply actions

It would be VERY interesting...

to see how the Baron/Monta plus-minus broke down by the half.

by John Raffo on Oct 15, 2009 11:54 AM PDT up reply actions

The only +/- I care about is

The team’s total, not the individual!

by banandy on Oct 15, 2009 10:30 AM PDT reply actions

And the Big Fat W, First time ever.

Everything starts out New, Gets Old and Dies or is Destroyed.

by HVYDRT007 on Oct 15, 2009 10:53 AM PDT up reply actions

Well of course

But that’s not really the point. People use stats to determine individual contributions from players in order to play arm-char GM/coach and figure out the best players to start, draft, trade for, sign, etc. At the end of the day, nobody cares if the stats are pretty so long as it results in a W. But ignoring the statistics in favor of simply looking at end result is troublesome. Is Robert Horry the best basketball player in history? Of course not. Considering it being a preseason game, I find statistical analysis particularly useful—- and breaking down individual contributions is more meaningful than the end result (in terms of preseason.)

by Michael White on Oct 15, 2009 11:04 AM PDT up reply actions

Agree...

I’m not a stat guy at all. But I LOVE boxscores… especially after a satisfying win or a puzzling loss.
One of the most frustrating things, now that I no longer subscribe to a daily paper is the difficulty in sometimes finding boxscores on line. Have you tried to find them on LATimes.com? They’re not on the same page as the game story. They’re usually not even linked. And when you find them you can only pull up one at a time.
When you open a dead-tree newspaper, BOOM… there they are, all together. You can check the entire league in just a few minutes.

by John Raffo on Oct 15, 2009 12:03 PM PDT up reply actions

I'm not saying stats aren't important

I’m just not a fan of plus-minus for the reasons that Steve mentions above.

by banandy on Oct 15, 2009 12:18 PM PDT up reply actions

Here's an interesting commentary on the plus-minus debate regarding Durantula

article

Basically, plus-minus is not pure and obviously has some noise in the numbers. In comparison, FT% and assist-turnover ratio are much clearer measures of individual performance.

However, I do think plus-minus could be much more useful when looked at for specific 5-man lineups instead of drilling down to the individual.

by banandy on Oct 15, 2009 12:38 PM PDT up reply actions

Good note, banandy

I was going to mention this. HAbbott at Truehoop worked on this issue for a couple of days, and in his “Letter to a Young Baller (Durant)” he broke down some of the things that Durant could work on to become a much better player—and help out his plus/minus.

by citizen zhiv on Oct 15, 2009 9:30 PM PDT up reply actions

Suggestion (ideas I also put down in a different post)

Why don’t statistitians use plus/minus primarily as a way to compare unit vs unit. That is, our starters vs. theirs, and our subs vs theirs. There are many times when the units are too mixed, with starters and subs all together, for this to work. But there are stretches when the starters are all in there, and then other times when each team has at least 3 (or 4, if you like) subs playing. If you made this sort of comparison, you could then tease out individual contributions on particular squads. For instance, we could find the plus/minus on “The 2nd squad when Thorton’s playing vs their 2nd squad.” We could compare this then with “The 2nd squad when he isn’t playing, vs. theirs.” This would gain us more precise insight into his particular contribution.

by silverman on Oct 15, 2009 1:52 PM PDT reply actions

This moneyball stuff...

If you’re interested, the Mavs stats’ guru has a very good blog about this stuff…he was interviewed recently about which lineups are better than others for different teams around the NBA…it was very insightful.

by banandy on Oct 15, 2009 3:47 PM PDT up reply actions

Enjoy

Good reads

by banandy on Oct 15, 2009 3:58 PM PDT up reply actions

Really interesting stuff...The Durant plus-minus argument is compelling...

The Durant plus-minus argument is fascinating… I haven’t seen enough of the guy but aren’t the conclusions pretty damning? Isn’t he just a not-good defensive player and a somewhat selfish offensive player?

by John Raffo on Oct 15, 2009 4:19 PM PDT up reply actions

Very informative... and obviously not a "daily" sort of analysis

I’m going to echo swamigusto’s comment below that it would nice to have a boxscore stat that provides more value than what we have now with plus/minus numbers. What the Mavs stats guru has done certainly has a lot of value, but it’s also much too robust for daily consumption. What I have in mind is a happy medium. If I hear that our 2nd squad went minus 12 against the Spurs 2nd squad (say, when at least 3 subs in for both teams), but that while Thornton was subbing our 2nd squad went only minus 5, I end up thinking that Thornton had a good game (and likely he did!). On the other hand, if I just hear that Thornton went minus 5, with no qualification, then I reach the opposite conclusion. In other words, an extra bit of analysis, which would be do-able on an interactive sports site, in some good measures corrects many plus/minus distortions.

by silverman on Oct 15, 2009 6:10 PM PDT up reply actions

front court units vs back court units

is a good unit division.

hockey is religious about units due to the in-game line changes.

but it works either way as long as you dont have a guy getting into foul trouble every night, haha.

No gang...... Independent....

by Clipochistic on Oct 16, 2009 9:50 AM PDT up reply actions

You could do all that...

But not within the context of a standard boxscore. I know I’m narrowing the argument but I suppose the next question is, is there a better individual player game stat that might be more valuable than plus-minus that also may be easily viewable in a boxscore? FT% and A/T ratio are easily derivative of numbers already there, so I wouldn’t include those.

by John Raffo on Oct 15, 2009 2:10 PM PDT reply actions

Perhaps not a standard box score...

but a sports page could give the basic plus/minus for how the squads match up, and then with a little work, you could create an new sort of interactive box score, say on ESPN, where we might request more specifics. It’s true, this won’t happen tomorrow, but with so many stats geeks out there, and some good programmers too, I say that it’s realistic.
A/T ratios and the like are good as well, but I’d really like to get at a players performance relative to the squad he’s in. If Thorton’s playing with lousy side-kicks, then a poor A/T ratio still doesn’t accurately portray his skills. I suspect most other stats we might imagine will have this same short coming.

by silverman on Oct 15, 2009 3:15 PM PDT up reply actions

Yes...

This might be the advantage of a web-based “boxscore” over the static newsprint kind. Some kind of an “interactive boxscore”. But I fear my brain is too small and inflexible for such a thing.

by John Raffo on Oct 15, 2009 4:09 PM PDT up reply actions

It's an interesting point

It’s funny that citizen swamigusto is lamenting his loss of the newspaper box score, while asking for more data. The newspaper page might have been able to place them all together, which I admit is nice, but the column inches are fixed, and adding advanced stats isn’t a good option. On the web, on the other hand, column inches are essentially infinite, and so it’s interesting that there is NOT currently (that I know of) a hyper-advanced box score available. Even places that trade in advanced stats don’t do them on a per game basis, as far as I know. Which is strange – the calculation has been done, presumably the objects exist to spit out the numbers – it really would be no work at all to include these things on a drill down.

So why don’t we have a PER column, next to a Win Share column, next to an adjusted plus/minus column, next to eFG, next to TS%, etc. etc.? Maybe I should do that. Hmmmm.

Or maybe it’s out there somewhere and we’re just missing it. Anyone know? John R?

By the way, popcornmachine.net, if you’re not familiar with it, has some very nice info that you can’t really get elsewhere. It’s the best place to look at the performance of units on a game by game basis.

In this world, you must be oh so smart or oh so pleasant. Well, for years I was smart. I recommend pleasant. - Elwood P. Dowd

by Steve Perrin on Oct 15, 2009 10:58 PM PDT up reply actions

I'm glad I'm not crazy

Thanks for your thoughts, Steve. I’ll check out popcornmachine when I get a second here. There are a lot of things that would be easy to do if someone just got around to doing them. I happen to think squad comparison is a great idea, for reasons I’ve already explained. If stats geeks can keep up with how the score varies for each individual playing, then I’m sure they can compile data in other arrangements.

It would be great if you could pursue this. I’m going to look around myself— I’ll let you know.

by silverman on Oct 16, 2009 10:07 AM PDT up reply actions

My point was...

…that I just noticed the plus-minus statistic in the standard box score and wondered at it’s value. And, yes, I do miss the big newspaper page that had all the data in it. But as the argument evolved here, I was also trying to chase some bigger ideas. (The Durant/True Hoop argument is fascinating and I missed it). But the idea of a web-based game-by-game reference source is a great idea… as long as it’s accessible and, at its top level, as easy to understand as a boxscore. So, I want both more… and less. I guess that’s anachronistic. So what?

by John Raffo on Oct 16, 2009 10:31 AM PDT up reply actions

Your point is a good one...

…that the plus/minus numbers are of limited value. I understand, too, you want something simple and accessible. But since you follow the Durant arguments, I’m sure you’d be able to follow what I have in mind. I picture getting to the ESPN box score, checking out the squad plus/minus stats, and then typing a name you’d like to focus on (or omit). “Give me the plus/minus on our 2nd team when ______ is in the game.”

My point: it could be designed in a very accessible fashion. Further point: If you don’t do this, then you’re stuck with boxscore stats, which have significant limitations. I don’t think what I’m suggesting is either too radical or too difficult.

I’m carrying on about this, really, because I’d so much like to see it.

by silverman on Oct 16, 2009 2:34 PM PDT up reply actions