Football analysis question

Everything else
1

OK, a question for all the (american) football geeks on this board.

It seems to me that hard football analysis is in the stone age in some ways, equivalent to pre-Bill James baseball.

There’s all kinds of analyses that I’d like to see. But here’s the one that frustrates me the most…

When people say “Team X has a great offense”, they generally refer to one of two things - scoring offense, or yardage offense.

But that’s really simplistic. A team that heaves the ball deep all the time will score more, simply because their average possession time will be shorte, Incomplete passes eat up no time, so some quick 3 and outs, and also some quick strike touchdowns, but no long drives. Shorter possessions means they’ll also get MORE possessions in a 60 minute football game. So much of their increased scoring may be a simple result of the extra possessions they get. But their defense will also have to defend more possessions.

Even worse, scoring offense stats include points score by the defense and returns teams.

What should really matter in evaluating an offense (or defense) is average points scored (or yielded) per possession. A team that scores a touchdown 60% of the time on offense is a great great offense, whether they take 2 minutes or 7 minutes a drive to grind it out.

Ideally, this stat would be modified in various appropriate ways - adjust for the average field position they start with (i.e. an offense paired with a defense that nabs a lot of takeaways or that has a great special teams return average may start off in great field position more often). You might also adjust in some way for end-of-half/game possessions (i.e. if you get the ball back 30 seconds before the end of the half, you shouldn’t be ‘charged’ a full possession). Finally, the stat would be adjusted for points yielded via turnover (if you throw an INT that is returned for a TD, you’re credited with -7 points), and probably even for net field position changes (driving from your own 1 to the opps 1, even if you don’t score, is a much more positive result than 3 and out).

So, is there any site/book that does this kind of geeky analysis? I’ve seen fotball outsiders and like it, but it tends to be focused on game situation analysis, not adjusted team comparisons.

2

I recommend you visit Football Outsiders and read everything there. You can’t do Bill James style analysis because football has so many confounding variables that any type of real analysis is nearly impossible. Trying to do a proper correlation between X performance and Y variable is borderline meaningless since there are like 80 other factors at paly.

What should really matter in evaluating an offense (or defense) is average points scored (or yielded) per possession.

Yes. When I was revisiting quarterback ratings I came up with the “Hook Quarterback Value” which basically comes down to one thing – how many points did that QB get on the board per drive, and how much passing did he take to get there? That’s all that matters in the end.

FO is going to be your best bet for this kind of stuff, and you should definitely check out their books (Pro Football Prospectus).

3

I second FO. They’re basically state of the art as far as that sort of thing goes.

4

Thorn and Palmer did the Hidden Game of Football. No idea how good it is, flipped through it at Barnes and Noble one day and found it kind of out there.

Football is the most old boys club of the major sports, I doubt there’s much room for new ideas. It’s also the most team-oriented and full of situations where teams don’t play to maximise total points, so statistical analysis is hard. No motivation+difficult task=Tom Brady has a QB rating of 92.3!

The backlash against that guy at Berkley who demonstrated the (IMO) obvious fact that football coaches are insanely conservative when it comes to making 4th down decisions shows that the current management doesn’t dig on new ideas.

Anyway, just in your analysis, scoring per drive isn’t the goal. The goal is to score more than the other team. A team that takes 7 minutes to score will have a devil of a time mounting comebacks. I’m hugely unwilling to do the math, but I imagine good teams want lots of possessions per game while bad teams want few(under the theory that each possession is another opportunity to leverage an edge).

5

FO’s work is based off of the Hidden Game. I’ve been trying to get that damned book for a year now with no luck. Amazon claimed to have it back in late October, but they never ever shipped it - I’ve still got my order in just for kicks.

6

I would disagree with the premise of this thread and say that baseball has actually just started to catch up with football in terms of individual analysis. Computer research, film, exhaustive psychological analysis, and coaches that work twenty hour days have all been a part of the NFL for decades. If you think all football players do is suit up and go bash it out, watch some of the behind the scenes stuff on ESPN, Inside the NFL, etc. It is mind-boggingly complicated.

The only reason there is not a Bill James-type in the NFL is that football is much more dependent on film study than stats and that is not available to the average fan. Pretty much all coaches and players do throughout the season (and off season) is study film. Obsessively.

7

Well, I realize that football management most likely uses more sophisticated analyses than the press/general public, but we don’t see those numbers very often. We’re stuck with the equivalent of baseball’s simplistic batting average - it gives us a rough, but incomplete and somewhat misleading look at the game. Purely looking at points for and points against is similarly crude when analyzing an offense or defense.

And I’m not entirely convinced that football management is adhering to rational analysis, either. Given how poorly they handle certain tactical situations (going for it on 4th down, etc), I’m not convinced that they handle other data all that well. In baseball, it took until, what, 10 years ago or less before people started realizing that OBP and Slugging were more important that BA. I suspect we’ll see a statistical revolution in football in coming years too.

Certainly, it’s more difficult in football to isolate the contribution of a single player. But analyzing an entire unit (offense, defense, special teams), is highly feasible, but I haven’t seen such analyses advance beyond the crude measures we had when I was a kid in the 70s.

8

Stats Inc. has a monopoly on those stats, and charges at least $50K to peek at even a subset. They have the data necessary to do a DB query like “What is Edgerrin James average yards/carry on 3rd down with Harrison in the game, Manning in the shotgun, and facing a nickel defense when it’s sunny outside?”

And I’m not entirely convinced that football management is adhering to rational analysis, either. Given how poorly they handle certain tactical situations (going for it on 4th down, etc), I’m not convinced that they handle other data all that well.

Head coaches are pussies. It’s that simple. Very often they play not to lose, not to win, which is why you see stupid shit like punting on 4th and 1 at your own 45 yard line when you’re down 21-7 with five minues to go. Or kicking a field goal from 52 yards instead of going for it on 4th and 4 when you still need 14 points to tie.

As Gregg Easterbrook says, when a coach goes for it he’s putting the responsibility on himself for making the call and he’s challenging his players. If he doesn’t go for it, he’s playing it safe and avoiding being second guessed by the media – he’s covering his own ass and trying to minimize the margin of defeat.

I suspect we’ll see a statistical revolution in football in coming years too.

Again, I think this is almost impossible due to the interlocking nature of everything. There are SO many real variables at play that you can’t just say come up with a stat that identifies one key area. If a running back is particularly successful running to the right of the line, is this because he has an affinity for that direction? Because his OL is strong to that side? Because his lead blocker blocks better that way? Because he’s played a lot of teams with shitty LBs on the perimeter? Because he has a tight end on that side that repeatedly drags coverage?

Certainly, it’s more difficult in football to isolate the contribution of a single player. But analyzing an entire unit (offense, defense, special teams), is highly feasible, but I haven’t seen such analyses advance beyond the crude measures we had when I was a kid in the 70s.

If you want to just quantify, sure, I think that can be done, but that is not interesting. What you really want is to come up with correlating factors so you can figure out why a particular offense is so great, beyond the “They get the highest number of yards, on average, on first down”. Okay, so WHY do they get that? Is it the players, play selection, opponents…?

9

Bacon is very much right, yes there are a lot of statisti

10

Again, I think this is almost impossible due to the interlocking nature of everything. There are SO many real variables at play that you can’t just say come up with a stat that identifies one key area. If a running back is particularly successful running to the right of the line, is this because he has an affinity for that direction? Because his OL is strong to that side? Because his lead blocker blocks better that way? Because he’s played a lot of teams with shitty LBs on the perimeter? Because he has a tight end on that side that repeatedly drags coverage?

Bacon is very right here, there are a lot more variables at play here when you’re talking about a football game then baseball - statistically it’s a lot harder to analyze. Part of the issue is that the public just doesn’t have easy access to these stats; not only do services offer them at a premium, but it’s hard to get a starting point for the regular joe. Do you know every single offensive and defensive formation on every single play? Or who’s playing? How can you keep track of this?

In football, you have the interaction of 22 players in every play. In baseball, you have the interaction of essentially 2 players at any one point. Statistically this is huge and is easy to analyze (in comparison).

I think the proliferation of stats in baseball is because it’s easy to do, essentially for anyone.

As for football coaches being pussies, you also have to realize it’s easier being conservative in football because there are so many fewer games. Every game is important and everything hangs on the balance. It’s the only major sport with the least amount of games - college football worse than that. A 10-6 team can get into the playoffs and a 9-7 team (ie. the Cowboys) will get out. That’s only one friggin’ game, and many of those games were decided on one play.

— Alan

11

Well, Troy Aikman redoes offensive stats and defensive stats to account for things like turnovers, scoring, conversions, etc. This is way better than the NFL’s total yardage approach. I’m sure you can find his formulae online somewhere.

There is a lot of statistical analysis of football going on out there, however. But in football, as BTG says, it just isn’t overly important because it’s even more a team game than baseball. A player’s contributions to the goal of winning just can’t be measured very well by stats. They’re nice for fantasy football (which I very much enjoy), but that’s about it. The media WAY overplays them, and so do some players. Peyton Manning didn’t seem too pleased with sharing the MVP with McNair because Manning had much better stats. But McNair was clearly an MVP that year. He pratically willed his team into the playoffs. Brady is another QB who is better than his stats (although his stats are pretty good this year).

12

Here you go, Phil:

http://snap.stats.com/stats/nflinfo/index.asp

13

Robert - that’s a good site - reasonably thorough individual stats, though a bit sparse on team stats (just the basics).

To the others - again, I realize that individual stats in football carry less weight than in baseball. But you can certainly do offense/defense/special team stats, and they’re entirely valid, and I think there are much better ways to parse the data than are currently done (at least to any widespread effect). I’m not sure if you’d have to go all the way to Stats Inc $50K package to do it - is there any source of play by play, or even drive by drive, data available for free/cheap electronically?

And yes, I realize that this would not necessarily be useful in a tactical sense (i.e. the Bengals are weak on screens to the left, so we’ll try a couple of those). But just getting a better understanding of the respective and relative merits of each teams 3 major units could lead to better decision making. For instance, a team who’s offense looks good based on points scored could actually be much worse than it appears, if it is benefitting from points scored by it’s defense, and giving up lots of points and turnovers to the opponents defense, and/or just generating a lot of points because of a quick playing style that leads to a lot of possessions.

There’s lots of related questions I have as a fan, and that as a GM I’d definitely have.

How much does ball control/time of possession affect late game defensive performance (i.e. do defenses truly tire faster than offenses, to a statistically determinable extent)?

How important are special teams? Not just “very important”, but how much does the deviation between having a top-5 special teams unit versus a bottom-5 unit affect winning percentage - gimme something concrete.

Do you really need to “establish the run”? If a team doesn’t run much in the first quarter, how does that affect performance later on?

Should teams be trying on-side kicks much more often in non-late game situations (i.e. surprise on-side kicks)?

I know there’s a lot of conventional wisdom on these and other topics. The C.W. may in fact be right. But given how often the C.W. seems wrong on topics that ARE rigorously analyzed, I strongly suspect that good analysis would shoot down some old theories (though not all of them).

14

nfl.com, espn, CNNSI, etc. all have play-by-plays, but you’ll have to write a scraper/parser (which I’ve mostly done). Unfortunately it’s not “well structured” data, so you can’t just create a grammar and have at it, a lot of the data is semi-random in how they enter it, so there’s a lot of special casing.

For those wondering, the reason I know all this shit is because I’ve been researching writing a football management game for a year now.

As for your other questions, I have seen a lot of different analysis, but it’s never structured. Usually it’s some site or author asking “Well, how much does special teams make a difference?” and they do the math and regression and come up with important info. FO has done a lot of this type of stuff in the past.

15

BTW, the football outsiders got to go meet Ron Jaworski and the NFL Matchup team. Now THAT makes me fucking jealous.

16

No shit. Jaws is the best pro football analyst on ESPN.

17

I hate Jaws’s analysis. I like Hoge better, actually. Of course, it may just be that I hate Jaws himself, but his film reports always seem so obvious to me. “Jake Plummer is a better QB this year because he makes better decisions.” Hmmm…really?

18

Yeah, but then he shows you specific instances. I mean, sure, any clown can say that, but he’s giving you the actual tape.

19

Yeah, I do love NFL matchup just because you get to see the coaches film. They really should show stuff like that on various websites and/or the NFL channel or something. Matchup is just about the only place you really get to see it like that.