Can someone explain the "plus-minus" system?

[Global Celtic]

I never figured out the "+/-" stats. What does it mean? How's it calculated?

I've heard it's a way to describe a player's impact on the result while he's on the floor. Is it that simple?

[Brickowski]

It is simple.  How did the team do when you were on the floor?  If you enter the game with the score tied, and leave when your team is ahead by 10, you are +10 .  If you play all 48 minutes and your team loses by five, you are -5 for that game.  And so on.  You can also compute the aggregate plus minus.

They use the same stat in hockey, I believe.

[Fafnir]

I never figured out the "+/-" stats. What does it mean? How's it calculated?

I've heard it's a way to describe a player's impact on the result while he's on the floor. Is it that simple?

It is that simple. There is a variant called adjusted plus/minus. What that does is it tries to control for the quality of team mates that a player plays with.

For example if backup player A only is in the game with Pierce/KG/Allen/Rasheed odds are he's going have a positive plus/minus. But that doesn't mean he's very good, just that he played with a strong 5 man unit.

Its a very noisy statistical measure, but its good for determining how much a player's defense, hustle, passing, and "intangibles" really effect the game. You just have to use several years of data to get an accurate number.

(FYI in adjusted plus/minus KG has been the best overall player of the decade, because of his defense)

[Fan from VT]

Yep, and then there's "Adjusted +/-." This is similar, except it tries to account for good teammates.

For example, let's say Ray is injured, so Tony Allen starts. All 5 starters play 48 minutes and win by 15. Therefore, Rondo, Tony, Paul, KG, and Perk all have a +/- of 15. But then Ray is healthy, so they go out and win the next 10 games by 20. So then Ray, Paul, Rondo, KG, and Perk would have a +/- per 48 minutes of about 20. Ray's "adjusted" +/- is +5 (his 20 minus his replacement's (Tony's) 15+/-). Tony's +/- per 48 minutes is still 15, but his adjusted +/- is -5, because the team is 15 points better than opponents with Tony on the floor but is 5 points worse than usual when he's on the floor.

That's obviously simplified, but that's my basic understanding. To calculate it for the season, they just tabulate the team's +/- when player x isn't on the court and tabulate it when player x is on the court for the whole season, divide by 48 minutes, and get an adjusted +/- per 48 minutes.



Anyway, that's why +/- is not useful in small samples. From my understanding you need a few seasons of data and for players to switch teams to get an accurate +/- rating. Otherwise, it just tells you who is better than whom on a given team; i.e. it could tell you that the celtics last year were best last year with KG at pf, second best with davis, and third with Scal. So you could assume that KG has the most positive impact on winning, Davis second, and scal third, so then you might conclude that KG is better than Davis is better than scal. But it doesn't tell you much about players from different teams. After all, KG is pretty darn good, so he's going to make his replacement look worse than if you are backing up, say, Verajao in Cleveland. That's why it takesa few years and some switching of teams to really figure out some useful data.

[LB3533]

Plus/Minus doesn't actually describe or represent the impact of a player.

You can do absolutely nothing on the court and gain a +20 rating solely on the work and hustle of your teammates.

Likewise, you can do absolutely everything for your team, within that stint you are playing, and register a -15 rating, but because your team is getting blown out it won't really show up that you actually were doing "everything" for your squad.

[Fan from VT]

Plus/Minus doesn't actually describe or represent the impact of a player.

You can do absolutely nothing on the court and gain a +20 rating solely on the work and hustle of your teammates.

Likewise, you can do absolutely everything for your team, within that stint you are playing, and register a -15 rating, but because your team is getting blown out it won't really show up that you actually were doing "everything" for your squad.

Right, and that's the idea of "adjusted +/-". Because if your team is that good but you stink, then over the course of the season, all the various combos will start to weed out which individuals are on the floor during the best stints, because there aren't 5 man units that play together, there's a lot of mixing and matching. So for your examples, the 4 good players would end up at +20 or so and the bad 5th starter would end around +12, around the rest of the backups or something. Still better than 0 because all the bench are playing with good players on a good team but not as good as the stars. The same is true on reverse for the one on a bad team. sure he ends up at -15 because it's a bad team, but everyone else will end at -25 or something, so his adjusted +/- would be +10.

Still not perfect, but better than +/- and it does take a long time.

[Fan from VT]

I think it's less useful for trying to compare two players from different teams. but i do think it is useful for starting to judge when a young player is better than a veteran starter but the veteran is still starting due to seniority/reputation. Helps dispell some conventional wisdom.

[Master Po]

Good explanations from Fan VT and others......

unless you are a coach or a scout and you want to use this "measurement" to combine with other many statistical yardsticks and on-court observations then I see this as a worthless stat - especially for 99% of fans.

but thank for further explanations and the question

[Fan from VT]

Good explanations from Fan VT and others......

unless you are a coach or a scout and you want to use this "measurement" to combine with other many statistical yardsticks and on-court observations then I see this as a worthless stat - especially for 99% of fans.

but thank for further explanations and the question

Yeah i think it is hard to use, especially with what is available to the public. But it does help provide evidence for why players like Bowen, Scal, and Battier seem useful and seems like good things happen when they're on the court, but you don't see "good" numbers.

[ToppersBsktball10]

Kinda cool stat, but doesn't really help evaluating a player.

[Fafnir]

Kinda cool stat, but doesn't really help evaluating a player.

Why do you say that?

[ToppersBsktball10]

Kinda cool stat, but doesn't really help evaluating a player.

Why do you say that?

Well it does help a little, but it's not something you can rely on. Should I further explain?

[Fafnir]

Kinda cool stat, but doesn't really help evaluating a player.

Why do you say that?

Well it does help a little, but it's not something you can rely on. Should I further explain?

What do you mean by rely on? Adjusted plus/minus has its flaws, but its a very good statistical tool.

[jdpapa3]

Adjusted +/- is one of my favorite stats out there...but I feel you are hurt when you have a good backup...I bet Boozer's stats in Utah were hurt by this since he had Millsap backing him up.

[Fafnir]

Adjusted +/- is one of my favorite stats out there...but I feel you are hurt when you have a good backup...I bet Boozer's stats in Utah were hurt by this since he had Millsap backing him up.

Adjusted plus/minus is supposed to correct for that. I haven't heard that's one of the problems, just that its a very noisy measure.