Teams play much, much better when behind

[TeamEd]

I've been looking at situational player tendencies in comebacks and in looking at league averages I noticed something really bizarre.
Basically, NBA teams all play much, much better when they're behind in score. I can't think of a reason that accounts for the extent to which they all play better from behind. I wonder if anyone can help me figure it out.

This chart is showing teams ordered from left to right by their overall +/- per48.* The high-low bars show the teams +/- while leading and while trailing. The red is +/- per48 while behind, the green is the same while ahead.

WTF?

Why are teams ridiculously better when they're behind. What is going on here?

I mean, correct me if I'm wrong, but this shouldn't happen. Like, I'm not surprised that there's some tendency for teams to play worse overall with leads -- especially big leads. You can think of garbage time, letting off the gas, bench lineups vs. starters, not wanting to run up the score. All that. But, this is a difference of ~25 +/- per48 across that board. That's huge!

Take the Bucks: When they trail, they play better than the Warriors do overall by a mile. When they lead, they play worse than the 76ers, again by a mile.

If basketball scoring is supposed to be a weighted random walk, this just doesn't make sense. So again, WTF is going on here?

* This is basically NetRtg per48 instead of per100 possessions. I can't calculate situational +/- per 100 possessions without per possession data, and I don't have that. The numbers are from NBA.com's in game splits.

[blabla]

Originally brought to the public by Matt Goldman et. al (SSAC), then further researched here
http://apbr.org/metrics/viewtopic.php?f ... of+leading

[Chicago76]

Don't know if you dug through the link provided by the poster above, but the fairly intuitive answer is that the object of the same is to win, which doesn't necessarily require you to outscore your opponent through the end of the game provided you have a cushion.

To maximize your chances of winning, with a lead, the goal is to maximize your win probability, which means reducing some of the variance around the outcome. Teams don't care if they win by 15 or 22 as long as they win, so if they're up 15 with 8 min left, they will look to slow the game down on the offensive end, which results in poorer shots for your O, but less time on the clock. Clock becomes more important than pts scored at times. It's kind of like a bend but don't break defense in the NFL--the idea there being that giving up yards is preferable to giving up a big play.

[BudTugly]

Referees.

[magicmerl]

It's clearly because the team that is ahead is playing for different things. Here are some ways that the team that is up could be not playing as good a basketball:

1. They are running out the clock on possessions, ensuring that they win, even if they get outscored the rest of the game
2. They rest their best players with the game wrapped up, helping prevent needless injuries

You also see this at the end of the regular season. Just like the final 3 minutes of a game is meaningless when one team is up by 20, whole games don't matter when playoff seeds are already locked up. Being healthy for the playoffs is overwhelmingly more important.

[colts18]

It's clearly because the team that is ahead is playing for different things. Here are some ways that the team that is up could be not playing as good a basketball:

1. They are running out the clock on possessions, ensuring that they win, even if they get outscored the rest of the game
2. They rest their best players with the game wrapped up, helping prevent needless injuries

You also see this at the end of the regular season. Just like the final 3 minutes of a game is meaningless when one team is up by 20, whole games don't matter when playoff seeds are already locked up. Being healthy for the playoffs is overwhelmingly more important.

The effect is still the same in the 1st half when teams are not resting guys.

[TeamEd]

I'm going to point out something I've missed here.

That's that the numbers get skewed as the lead changes. Every time the trailing team takes the lead, they'll record +1 or +2 while trailing, but if they then lose that lead they don't get a corresponding -1 or -2. This happens because basketball scores largely change by 2 or 3 points at a time.

This has the effect of racking up +s on the trailing side and -s on the leading side.

Does it cover everything? I don't know, you can see that teams that will tend to regain leads often are logically more likely to be good teams. On the chart, you can see there is a trend in that graph where the better teams have better trailing +/-s and the worst teams have worse leading +/-s.

As a result, this line of inquiry into ahead/behind play is essentially dead. There might be other ways to look into this, but I can't do it with +/- numbers.

[165bows]

Might be covered in the links above, but I've read similar analysis that says that teams generally take greater risks (ie, more threes) when behind as compared to ahead, where they tended to get a lot more conservative.