## Interceptions and Counter-Momentum

Over at the Community Site, contributor Andy Steiner has a very intriguing post. I normally link to the community posts in my weekly roundup, but this one was so interesting it merited some extra attention. Andy looks at whether teams tend to score more points following an interception than when they get the ball by other means.

I'd like to know what everyone thinks. I've been meaning to do a grand study of momentum, but haven't set aside the time to do it justice. Any suggestions on how to structure further research, using EP or WP perhaps, are appreciated.

Thanks, Andy.

### 14 Responses to “Interceptions and Counter-Momentum”

1. Phil Birnbaum says:

Andy says he doesn't know how to compute statistical significance for this. The easiest way is this:

Since the interception sample had 501 data points, choose 501 datapoints randomly from the non-interception sample. Calculate the curve from the "fake" sample the same way as the original sample.

Repeat 999 more times. Then, see how many fake samples were at least as extreme as the real one. That's your estimate of the significance level.

2. Jeff Clarke says:

Very good work.

I tend to think that the counter momentum is probably just a sample size anomaly. I'm not sure why you'd do worse after an interception. Perhaps there is something to the fact that the offense isn't ready to get the ball yet. They didn't have enough time to meet. But then again, that would mean the opposing defense didn't have enough time to rest.

I definitely don't believe in sports momentum in general. I think that its an offshoot of the fact that people tend to see patterns in things that are just created by pure randomness. Of course, you sometimes get a bunch of good plays in a row but just as often you seem to hear about a "momentum shifter". If momentum was as valuable as it seemed, it wouldn't be shirting to the other side as often.

Here is an interesting little parlor trick.

Open Excel and goto A2. Write "=RAND()-.5+A1". Copy it to cells A3:A100. Make a graph of the whole series. What do you see? Well from the confines of the experiment, it should be obvious that you have a line thats next move is based on pure randomness, but I bet you see some identifiable patterns anyway.

Given enough time to explain it, I'm sure you can see some way to define "momentum" in the line and ways to predict when it will change.

Hit F9 a couple times and you'll see a whole new set of patterns.

I think that most people make too much out of momentum because they instinctually think that random numbers should not form patterns, but random numbers do form patterns. They are just unpredictable ones that can reverse themselves at any moment.

I admit that I used to buy into sports momentum. When I play basketball and hit 5 shots in a row, I sure feel like I'm "hot". I'm not. Its just randomness forming an unpredictable pattern.

3. Dave says:

I've wondered whether teams get a big play or score TDs more often on the first play after an interception than they do on first plays of drives from similar field positions. Might be an interesting study for an aspiring statistician.

4. Dave says:

Oh, and the reasons behind my hypothesis in the above comment are that a) I feel like I've observed this and b) perhaps having to unexpectedly play is more difficult on defense (where you are more reactive) than on offense.

5. Chris says:

Would the 5.3 be from a missed PAT? Similarly you might expect a couple of 2pt tries to appear in the data.

I'd be interested to see if using Brian's definition of 'normal' play - i.e. no more than a 10 point game, 1st/3rd qtr - would make any difference here - hard to know how many of these are perhaps not converted because the team making the INT is either far ahead (so doesn't need the score) or far behind (so is likely to be less capable of making the score) when the interception was made.

One final thing - it looks like this analysis doesn't include INTs returned for a TD... Wonder how you could assess the 'momentum' of those? Naïve assumption is that such plays would be responsible for the biggest 'momentum' (read EPA) shift of all.

Very interesting to see the analysis though - and certainly goes against my instinct (though stats often seem to!).

6. James says:

I agree it could be the surprise factor and it is not momentum. ( assuming it holds up as significant) therefore you would expect to see the same effect for fumbles but not for punts ko and missed fg. Also could he test gregg easterbrooks krumble hypothesis that a kr fumble is more likely to result in a score even when accounting for field position although I guess you would have to correct for the fact that high scoring teams kickoff more than low scoring teams
Ja

7. JMM says:

If Mo is put there, he could be hiding in the live probability charts. I don't know a numerical method, but there are clear shifts from a prolonged positive slope to a prolonged negative and vice versa. I also don't know if, after the fact and only relying on play by play accounts one can determine a "cause."

8. JMM says:

Not " put", "out there"

Thank you iPhone Autocorrect.

9. Ian Simcox says:

The reason people think there's momentum is because they see that teams score more often after an interception than a punt. But there's a simple explanation for that.

The average starting field position following an INT (from the 2005 PBP data) is around your own 45. Following a punt, it's your own 25. Those 20 yards are what give the appearance of 'momentum'.

Absolutely agree with Jeff Clarke - if momentum is real then why does it swing around so much?

10. Andy says:

Wow, thanks everyone for the comments. I will work on the statistical significance part as soon as i can, thanks Phil.
I did not even think about the missed extra point, i suppose that would make perfect sense!
About interception returns for touchdowns, that's correct, they weren't included. I do think, simply from watching games, that while TDINTs are huge in terms of perceived momentum, normal interceptions are also pretty large in terms of perceived momentum also, so i thought i would have seen some effect.
Also, this is similar to what the David Romer paper found, that there was actually a letdown after really good plays, so it is possible that this is a real effect (im not saying for sure that it is though).
Lastly, I agree Ian, much of the perceived momentum is probably due to a huge advantage in the game, which I'm not denying. Interceptions are huge, but there is no extra bonus; you have your reward in full by having the ball.

11. Brett says:

There is natural momentum in basketball because the probability of scoring is affected by the manner in which possession is gained. For example, steals and blocks lead to fast breaks, and made baskets force the other team into a half-court set.

Psychological momentum is certainly possible in football, but there is no natural momentum in the game because the ball is dead between plays and there is a completely different set of players on the field following a change of possession. The only situation where natural momentum is possible is when a team is running no-huddle offense and doesn't allow the defense to substitute. However, this is often negated by the fact that most teams become more mistake-prone when they don't huddle. The obvious exception is the Colts, who have shown the no-huddle offense to be an effective way of creating natural momentum without an increase in mistakes.

12. Ian Simcox says:

Andy - absolutely, there's no reason to deny that interceptions give you a massive advantage and it is indeed true that (in most cases) you are more likely to score after you gained possession with an interception than by a more common method such as kickoff or a punt.

So that fact is correct, and when the TV says that teams score more often after INTs they aren't lying. They are just attributing the increase in scoring to the INT momentum, when it's actually just that you start with good field position.

13. Michael Beuoy says:

I wonder if teams are more aggressive after getting a possession off an INT. Kind of like they're playing with "house money".

As a Colts fan, I've noticed that Manning likes to take a shot deep down the field on the first play after an interception.

14. Andy says:

Also, Chris T,
I already have limited the data to the first and third quarters, score within 10 points.