The merits of the often used statistic of player +/-

[bostonman1]

Referencing a player's +/-  happens often on this site when assessing a player's performance.  Many feel it is a valuable statistic while others (such as myself) feel it is vastly overrated.

I thought today's game against NY would be a perfect time to see just how insightful +/- is...

For anyone who wants to see the actual stats -
http://scores.espn.go.com/nba/boxscore?gameId=320304002

Take a look at the Celtic's starting 5 tonight -

Rondo had an historic triple-double and was +5 in his 48 minutes.
Pierce was an offensive force all night and hit the 3 pointer which gave the C's a chance, +3 in 46 minutes.
KG struggled from the floor but hit some big shots and played solid D, +10 in 41 minutes.
Bass scored well, had 4 huge blocks, +2 in 30 minutes.
Ray Allen had 4 steals and hit two huge shots in OT but for much of the game was invisible offensively.  However Ray was a +22 in 45 minutes!!! 

If someone had not watched the game and just looked at the +/- stats it would seem Ray was phenomenal and the key to today's victory.  In fact he scored only 12 points and in watching the actual game was clearly the team's least effective starter.

To recap tonight's highlighted examples -

Pierce was outstanding tonight and hit the key shot in giving his team a chance to win in OT. 
Rondo was dynamic in all phases tonight and had a game that will go down in the record books.
Ray hit a huge 3 in OT and converted a layup off a pass from Rondo right after.  Otherwise he had a marginal game.

Yet Ray had a +/- which dwarfed Paul, Rondo and anyone else on the team.  His +22 is over 4x Rondo's, over 7x Paul's and over twice his closest teammate KG.

Often when a player has a bad game the counter argument is 'well he couldn't have played too poorly since the team was +5 during his minutes.'  Highly doubt anyone would say that Ray had a game that benefitted the Celtics more than 4x what Rondo contributed.

No matter how great a game an individual player has, their +/- depends greatly on the other 4 players on the floor for those minutes. 

Tonight's game is the perfect example of how misleading +/- can be as a meaningful stat.  When the least effective starter has the overwhelming advantage in +/- it definitely calls into question it's validity...

[Kiorrik]

Nobody is saying +- shows you exactly how good someone played.

It's just another stat. Like rebounds.

Someone can get 0 rebounds and play a GREAT game.

Someone else might get 7, and play like utter utter carp.

+- is just another stat

[bostonpatriot]

Plus/minus for a single game (or even a dozen, or a season) is completely meaningless. I don't think many people actually argue what you are refuting. The statistic may be used often, but not in that way.

Adjusted +/- with a large sample size - a few seasons, ideally - means something.

[kgiessler]

Yeah, or could point out that in the 8 minutes Ray was not on the court, the team was a whopping -18, which shows just how impressive his +22 stat is.

For whatever reason, his presence on the court was clearly helping the Celtics. This is no slight disparity, which is a situation where +/- is meaningless, because small numbers are often the difference between a bounce going one way or the other.  

It may have been because it forced the NYK defense open to guard him.  Or it could have been that he was performing well on defense.  The only way it doesn't point to his performance as a player is it could have been that the lineups doc used in his absence were poorly constructed.

It may also mean that the Celtics would have done much better in this game if Pietrus had knocked down his shots, as it appears many of those misses came while Ray was sitting: http://popcornmachine.net/cgi-bin/gameflow.cgi?date=20120304&game=NYKBOS

What's odd, is that out of all of the stats in basketball, +/- is really the only one that matters.  If all of your players end up with a positive +/- for the night, your team wins...if the summation of your team's players is positive, your team win...thats just how it works.  Conversely, if either of those two is negative, your team loses.  Triple doubles, big scoring nights, rebounding numbers, none of them are as close to representing winning as +/- is.

[bostonpatriot]

Yeah, or could point out that in the 8 minutes Ray was not on the court, the team was a whopping -18, which shows just how impressive his +22 stat is.

For whatever reason, his presence on the court was clearly helping the Celtics.

As the OP says, a player can be hurting his team and still have a massively positive +/-. He just needs his teammates to play well or the opponents to play bad.

[kgiessler]

Yeah, or could point out that in the 8 minutes Ray was not on the court, the team was a whopping -18, which shows just how impressive his +22 stat is.

For whatever reason, his presence on the court was clearly helping the Celtics.

As the OP says, a player can be hurting his team and still have a massively positive +/-. He just needs his teammates to play well or the opponents to play bad.

Correct, and if that were the case, the team wouldn't have had performed so poorly when he was off the floor.

For instance, what you are saying might be valid in a situation where someone has a +/- of +5, and their team wins by 27.  In that case, while the player was on the bench, the team did better than while the player was on the floor.  This is not that case.

[LB3533]

A few nights ago during the ESPN broadcast for the OKC game, Jeff Van Gundy was saying that over the last 2 or 3 years...Nick Collison was like ranked 5th in +/- behind the likes of Lebron, Dirk, Wade etc.....

But yea...some guy named Durant and maybe Westbrook was or were more responsible for that high +/-, not Nick Collison.

Edit: Here is the gameflow chart for the Knicks/Celtics game: http://popcornmachine.net/cgi-bin/gameflow.cgi?date=20120304&game=NYKBOS

If you will notice, Ray's 2 positive stints during regulation (1st and 3rd quarters), he contributed 0 points and 0 positive auxiliary stats.

Both Rondo and Paul played the entire 3rd quarter and Paul contributed 15 points while Rondo had 7 dimes and 5 boards.

Ray was fortunate enough to be taken out near the end of the 3rd when the Knicks went on a 7 nothing run to close the quarter out. We had our largest lead of 15 but that 7 zip run by the Knicks trimmed our lead to 8 going into the 4th.

But yea, the majority of Ray's +/- came in that 3rd quarter when he didn't produce a heck of a lot....namely zero.

But the guy was sick and probably not 100% so you can't fault him. Also, our offense isn't the greatest for Ray since we lost 2 of our best pick setters in Perk and BBD.

[bostonpatriot]

Yeah, or could point out that in the 8 minutes Ray was not on the court, the team was a whopping -18, which shows just how impressive his +22 stat is.

For whatever reason, his presence on the court was clearly helping the Celtics.

As the OP says, a player can be hurting his team and still have a massively positive +/-. He just needs his teammates to play well or the opponents to play bad.

Correct, and if that were the case, the team wouldn't have had performed so poorly when he was off the floor.

For instance, what you are saying might be valid in a situation where someone has a +/- of +5, and their team wins by 27.  In that case, while the player was on the bench, the team did better than while the player was on the floor.  This is not that case.

Why not? Maybe when he was off the bench the other 4 guys played worse and/or the opponent played better. It's just the difference between some made/missed shots.

Not, it's valid to any small sample. And that includes an entire season - let alone a single game. Not even the biggest +/- apologists use it in that way.

[Donoghus]

It's an over-rated stat.

[indeedproceed]

People who use +/- from a single game to draw any real conclusions don't understand its true value.

+/-, just like many other stats, gets a lot better with larger sample size (say, a whole season). Even then, there are outliers. If you're on a bad team, and you play a ton of minutes, you're gonna have a bad +/-. So, I find it useful to use some combination of +/- with On/Off stats. And even then, you can get outliers.

No one statistic tells a whole story. Some stats tell bigger parts of the story than others, but no one gets it completely right.

[bostonman1]

A few nights ago during the ESPN broadcast for the OKC game, Jeff Van Gundy was saying that over the last 2 or 3 years...Nick Collison was like ranked 5th in +/- behind the likes of Lebron, Dirk, Wade etc.....

But yea...some guy named Durant and maybe Westbrook was or were more responsible for that high +/-, not Nick Collison.

Saw that also, great reference point...

[Boris Badenov]

You can not - CAN NOT - use an example from one game to say anything about the value of a metric, whether it's this one or any other one.

The entire discipline of statistics is based on repeated observations of data. The more data you have, the more learn. One observation tells you nothing.

You could go and pick one game where the C's were outrebounded by 20 and won. Does that mean rebounds don't matter? If they shot 40% and won, does that mean that shooting percentage doesn't matter? Etc.

[GranTur]

Like all stats, you need to look at the trend of that statistic over time to get a real idea.

[bostonman1]

The +/- is much more relevant and useful when looking at it from a lineup standpoint. 

An individual player's +/- is influenced by so many other factors.  A great player could have consistently effective games but if he has 4 mediocre teammates his +/- will obviously be lowered.

Conversely a mediocre player could be the beneficiary of playing with 4 effective teammates on a winning team and his numbers will look outstanding even with minor contributions for that player.

[Q_FBE]

The theoritician in me is that Ray helped spread the floor and caused Rondo and Paul Pierce to have open driving lanes to do damage. He was battling a personal illness. He still scored some key buckets out there in OT and in the 3rd qtr as Boston built their double digit lead. Oh, and he took care of his defensive assignment quite nicely.

+/- stats must be considered in conjuntion with other metrics like points per shot, points/100 possesions played, defensive points allowed / 100 possessions played, defensive plays made (Def Rebounding + Steals + Charges Taken + Blocked Shots / 100 defensive possessions played) to get a complete picture of the players effectivity.