[spongerat]

turnovers are supposed to be random. then how can a team supposedly be good at "forcing turnovers"?

furthermore, if they ARE assumed to be random, how can one use turnover differential as a predictive stat when capping a game IF it is random because you have no idea what will happen? If they are random then shouldn't the differential revert to the mean?

[Pancho sanza]

I did a turnover study years ago.

Only real nugget i found is that if you were to compare the best/worst turnover teams after say the first 8 games of the year (or some other point) those teams tended to go the opposite way turnover wise the last 8 games of the year.

Makes sense in a way, teams with a lot of turnovers work on protecting the ball, eliminating turnovers while those with very few probably got some lucky bounces or start to get sloppy with the ball since they feel turnovers arent a problem.

Never was able to use it successfully betting wise.

[Dark Horse]

furthermore, if they ARE assumed to be random, how can one use turnover differential as a predictive stat when capping a game IF it is random because you have no idea what will happen? If they are random then shouldn't the differential revert to the mean?

That seems to be accurate (reverting to the mean as a way to forecast). Here's an article with some qualifiers.

http://www.twominutewarning.com/g13.htm

[Munson15]

I've read that there is an inverse effect as well. I wish there was a way to back test 2001 and forward.

[Justin7]

One approach I'd use... data regression. If a team averages "x" fumbles, or "y" interceptions, how many would you expect in the next game?

Same question for defense.

Easy test to do if you have the data.

The quick answer: the past is almost completely irrelevant. There is a stronger correlation between number of offensive plays vs fumbles than pats fumbles. Interceptions... I'd speculate the strongest predictor would be total passes thrown * average number of d-backs on other side. If you're always throwing into pass defenses (like bad teams in 3rd and long, or down a lot) you'll throw more picks.

[spongerat]

so it seems counterproductive when i see a capper use a formula that factors in turnover differential when calculating the score?

[Wheell]

Forced fumbles is predictive. Interceptions are also predictive (on both sides, thrown and forced).

There is a continuum in terms of pass defense. There are pass defenses that use man-zone hybrids to generate as many turnovers as possible, but this also imposes certain costs on the defense in terms of efficiency.

The best thing to do is try to break the game down into component stats and generate ratios: x tipped passes = y interceptions. That can allow you to tell which teams have been "lucky."

footballoutsiders can be helpful when doing this kind of research.

[Peep]

Never was able to use it successfully betting wise

I would think it would be useful in terms of prop bets. IF you know who is more or less likely to fumble, you know a bit about who will punt the most for example.

[Dark Horse]

We need better stats. If we could qualify an offensive line by how many seconds (1 or 2 decimals) they gave the QB, and how little time the QB needs to throw with accuracy, that might help. The difference between the two would be a safety cushion against picks. You could have such time stats for both the offensive and defensive lines. A blitz is a lot less successful against a QB with a quick release (not rushed). But a rushed QB is likely to throw picks.

Can you tell I'm reading up on horse racing?

[fiveteamer]

Darky if you are using math programs for horseracing, there are more random factors than football. Good luck with alllllll of that.

[Dark Horse]

Darky if you are using math programs for horseracing, there are more random factors than football. Good luck with alllllll of that.

Thanks. My luck is all digital.

I came across a speed figure that is more accurate than Beyer, so am reading up some more. May not go anywhere, but there are always ideas that carry over to other sports. Looking at it from the racetrack, football turnovers are often a function of (bad) timing. So why don't we measure the timing features involved? Players always say football is about split seconds. So let's found out which teams have the edge in split seconds.

========================

[waiverwire]

Thanks. My luck is all digital.

I came across a speed figure that is more accurate than Beyer, so am reading up some more. May not go anywhere, but there are always ideas that carry over to other sports. Looking at it from the racetrack, football turnovers are often a function of (bad) timing. So why don't we measure the timing features involved? Players always say football is about split seconds. So let's found out which teams have the edge in split seconds.

========================

Not sure if you're refering to Ragozin's figures, but they would appear to be a lot better than Beyer, and his book was a pretty big inspiration for me in building a statistical model for daily fantasy baseball contests.

[Dark Horse]

Carroll's figures.

In order to determine what type of timing edge a team has I need to come up with a mean figure for the league, for the time between the snap and the moment the ball leaves the QB's hands (including percentage of completions). The final number would contain the questions: how much time does a QB get? how much time does he need? how much of a cushion does that give him? And how does this cushion measure up against the league? I would need a digital timing setup, and would have to tape and review every game. A lot of effort for a number that, in the end, may or may not mean much. Then again, imagine being able to say that team A has a 0.78 second-per-snap edge over team B.

[Wrecktangle]

I see turnovers as mostly random, but if a team has a lot of TOs, the next week the line can move heavily against them and create an opportunity. Doesn't happen that often anymore it seems. I used to track this angle, but gave it up for black boxie stuff.