Weird Stat Question - too many overtimes?

[fluffernutter]

I have no idea how to look into this, or even what I am asking precisely.

I'll try to explain. My feeling is that there are WAY too many overtimes given the number of games played.

I don't think this is due to any referee conspiracy to get the games closer at the end, rather the human tendency to let up when you have a big lead and slow down on defense. Or MAYBE a vague combination of the two.

Is there any way to statistically define the likelihood of a given game ending in an overtime given team quality? And them look at the total number of games in a season and the number of overtimes and see if they are within a normal range.. out outside of it?

For example, if team A averages 150 possessions in a game and scores on 3pt/2pt/foul 20%/80%/20% of the time at rates 35%/50%/1 point per foul on average) and goes up against another team with different values, and you run a simulation on each 150 possessions and add up the numbers 1 million times, what percentage of the time is there a tie after 150 possessions, vs. real life? Rinse/repeat for all teams in the league.. I dunno does this make sense?

[blabla]

I can say that the effect of teams being down scoring better than expected is very real. See http://apbr.org/metrics/viewtopic.php?f=2&t=8501
Now, whether that's due to the refs, due to players not caring as much when being up, or players giving more effort when down, we don't really know

My gut says only minimal influence is from the refs and most of it is 'not caring' when up

Obviously you could simulate, but it's 180 possessions, not 150.. you'd have to account for teams having different strength and you should look up the exact distribution of 3s/2s/1s/0

[Knosh]

I have no idea how to look into this, or even what I am asking precisely.

I'll try to explain. My feeling is that there are WAY too many overtimes given the number of games played.

I don't think this is due to any referee conspiracy to get the games closer at the end, rather the human tendency to let up when you have a big lead and slow down on defense. Or MAYBE a vague combination of the two.

Is there any way to statistically define the likelihood of a given game ending in an overtime given team quality? And them look at the total number of games in a season and the number of overtimes and see if they are within a normal range.. out outside of it?

For example, if team A averages 150 possessions in a game and scores on 3pt/2pt/foul 20%/80%/20% of the time at rates 35%/50%/1 point per foul on average) and goes up against another team with different values, and you run a simulation on each 150 possessions and add up the numbers 1 million times, what percentage of the time is there a tie after 150 possessions, vs. real life? Rinse/repeat for all teams in the league.. I dunno does this make sense?

Check out http://apbr.org/metrics/viewtopic.php?f=2&t=8501

According to this, an away team that is down 10 points is expected to score ~3.5 points more PPP than if it were tied. Inversely, if they were up 10 they would be expected to score ~3.5 points less than if they were tied. Down 20 you're expected to score ~6.2 PPP more than 'normal' - that's like replacing an average offensive player with LeBron

The analysis isn't exactly geared towards OT games, but basically it says that games tend to be "close", because whoever trails gets a boost.

edit: Looks like blabla already posted it while I was digging up the link...

[fluffernutter]

That makes it extremely likely (nearly certain) that we see vastly more overtime games than we should if the team behind played at the same rate as normal, and the team ahead did the same.

As the paper states, the team behind gets better, and the team above gets worse.

How extremely odd. I mean from a psychological standpoint it makes sense, but the amount of change seems kinda wacky. That is a LOT.

I don't know how to run such a simulation myself (badly outlined in my initial post), as I am an idiot, but I'd love to see the results if somebody did even just a partial run.

[ElGee]

with regard to the Goldman study, it can be explained by: Regression to the mean. Human tendency to play the scoreboard (and exert less effort). And in sizable leads later in the game, the very real tendency to play weaker lineups with a comfortable lead.

But that doesn't explain if there are an abnormal number of OT's. What you're asking is, given the distribution of scores after 48 minutes, does the needle land on "0" (for net points) more than it should?

If we look at all games from 2007-2012, there were 351 OT games of 5,910 (5.9%). (This year's NBA had 6.4%, FWIW.) We can clearly see that the following game-outcomes were more common:

Score after reg -- % of games
7 6.7%
5 6.4%
8 6.4%
0 6.3%
9 5.7%
6 5.6%
10 5.2%
4 5.1%
3 5.1%
11 5.1%

7-point games are the most common in this sample. However, OT games are the only one of that group to fall more than a standard deviation away from the mean (by a basket). Without further statistical analysis, this suggests there is definitely some merit to the OP's observation. I would suggest we see abnormal numbers of OT games at the expense of 1 and 2-point games because the game does not truly end at a random time -- teams have awareness of the end of the game and enact high risk/high reward strategies when trailing in the final moments to extend the game.

[Knosh]

If we look at all games from 2007-2012, there were 351 OT games of 5,910 (5.9%). (This year's NBA had 6.4%, FWIW.) We can clearly see that the following game-outcomes were more common:

Score after reg -- % of games
7 6.7%
5 6.4%
8 6.4%
0 6.3%
9 5.7%
6 5.6%
10 5.2%
4 5.1%
3 5.1%
11 5.1%

Did some quick and dirty analysis along those lines. It looks like you looked at "absolute scoring margin", so for example Score=7 means home team wins by 7 or away team wins by 7.
This is how it looks:

edit: Sorry, pictures seem to be too big, you can view them in full by opening in a new tab...

(I did cut off huge blowouts)


If you treat home team wins by 7 and away team wins by 7 as two different events, it looks like this:

(Negative values means away team won, the big spike in the middle is indeed 0)

[blabla]

I think we see a lot less 1-3 point difference because the losing team, in the last 20-30s of the game, is *always* either trying to shoot as quickly as possible or foul. A lot of the quickly taken shots don't go in. The winning team thus scores more points than if both teams had continued to play without any change in strategy
If the game is tied with <24s left the defending team is probably trying to avoid fouls and just force a miss

[montestewart]

This paper asserts an apparent referee bias in favor of teams that are behind in a game, teams that are behind in a playoff series, and home teams. All of these biases, especially the first, could contribute to more OT games.
https://ir.library.oregonstate.edu/xmlu ... sequence=1
The paper speculatively connects these biases to profitable scenarios, equating closer games and series, and home team winners, with increased revenue. However, it does not assert that the biases are conscious.