Comments on Advanced NFL Stats: What Makes Teams Win? 2

@ Brian, You have a lot of patience with people ...

2010-08-06T07:57:07.149-04:00

@ Brian,

You have a lot of patience with people who obviously don't understand regression analysis, and who comment without bothering to look at the actual statistics. It is borderline insulting to your research lol. Either way, I can respect your tolerance.

Kick Return TDs may not be repeatable, but Kick Re...

2010-06-28T02:41:56.309-04:00

Kick Return TDs may not be repeatable, but Kick Return TDs Allowed are very repeatable. See: 2009 Steelers.

"Penalty rate is defined as penalty yards per...

2010-05-16T22:35:28.822-04:00

"Penalty rate is defined as penalty yards per play."

Is that offensive plays? Offensive + defensive plays? Including kicks and special teams? After all, penalties can come on any play on either side of the ball.

If it is all plays, then I can't see why penalties/play should be much different than penalties, since most games have a similar number of total plays. (Of course, one of the great things about stats is discovering things you would not otherwise have suspected. :-) )

If it is offensive plays, then it would seem that no matter how many penalties you have, you are less likely to win if you have fewer offensive plays. So the combined stat would simply be reconfirming that fact.

--------------------------------------------

As an extension to what you have done, I would think that separating out "offensive penalties per offensive play" and "defensive penalties per defensive play" would be informative. Then you would know if offensive or defensive penalties are more critical.

Oh, I totally plan on filching FO's stats. It's ju...

2007-07-13T14:29:00.000-04:00

Oh, I totally plan on filching FO's stats. It's just a shame they don't have weekly archives of DVOA beyond overall Off/Def/ST.

An another idea for a ST stat is how much field position is gained by the coverage/return unit on average. This would be a little less dependent on field position because you'd be counting touchbacks against the coverage unit. Punts don't usually go further than 40-50 yards, and team don't normally punt from within the opponents' 35-yard-line. So the pooch punts to pin down opponents near their own end zone shouldn't adversely affect such an average.

I like your theory on def penalties.

On ST,...

2007-07-12T22:38:00.000-04:00

I like your theory on def penalties.

On ST, I agree drive starting position is important, but it's not really dependent on ST play. Starting field position is overwhelmingly due to offensive or defensive play.

Consider two extreme examples. The game starts and your team kicks off. Your defense gets a 3 and out and your team gets the ball on your own 39.

Your offense gets 2 first downs but stalls at your opponent's 40. It's 3rd and 9 so you punt. Your opponent starts his drive on his own 15, or 20, or whatever.

In the second example, your team kicks off, and your opponent gets 3 1st downs, then punts. Now you're starting from your own 15.

See the swing. It's huge and has very little to do with special teams. They can affect things at the margin, so perhaps you're starting at your own 25 instead of your own 15. But that difference in field position is dwarfed by the direct effect of the offense and defense.

To put things in perspective, a single holding penalty causes a 1st and 20 (or could call back a long TD pass). The best kick return team in the history of the NFL gets you and extra couple yards per kick. I exaggerate a bit, but you see the point.

I agree that you can't dismiss ST. But the effect is very small and very difficult to measure. I guess that's the legwork you mentioned.

Maybe we could add FOs metrics into a comprehensive model to see if they're on to something.

On special teams: I don't think you can entirely d...

2007-07-12T22:22:00.000-04:00

On special teams: I don't think you can entirely dismiss the importance of special teams. Starting field position does directly affect a team's probability of scoring by a significant amount. The ST stats available in box scores are just really lossy information. To get a good special teams input, however, you'd need to do quite a bit of legwork (no pun intended). You'd need a metric of how much the return/coverage unit increased/decreased the probability of scoring on the subsequent drive. For that, you'd need information on every drive and every game for a few seasons at least to get some good idea of P(score | start at X yard line). And then you'd need the play-by-play data to figure out from where the ball was kicked and to where it was returned. I believe that's more or less how Football Outsiders does it. I don't mean to treat them like demigods, but they've obviously researched this more than I have.

On penalties: I was thinking about that maybe more defensive pass penalties are indicative of tighter coverage on WRs. Pass interference and illegal contact calls are always inconsistent (IND@NE 2003 AFC Champ.). If the DBs are mugging the receivers throughout the game, pass efficiency is going to go down, and the few penalties they get called for are acceptable losses. Or perhaps better pass defenses are able to cover recievers so tightly without major contact that they are targets for ticky-tack calls.

If you were able to break down the penalties by type and check correlations with pass def. eff., that might explain the positive correlation of defensive penalties to wins.