Comments on Advanced NFL Stats: Full Review of Game Theory Run-Pass Balance Study

In very simple terms, could you answer this questi...

2011-01-13T22:07:37.225-05:00

In very simple terms, could you answer this question. Does good running efficiency increase passing efficiency? I have read lots of your stuff but still couldn't seem to find an answer to this question.

Hi Brian, Had a question regarding this bit "...

2010-08-13T10:26:39.816-04:00

Hi Brian,

Had a question regarding this bit "For example, 0.60 WP is exactly twice as good as 0.30 WP" - are there not situations where external factors (such as Playoffs considerations) will affect this... ie if a team has 2 remaining games, and winning both guarantee it a Playoffs berth, but but losing either would deny it the berth... then 0.3*0.3 = 0.09 would be their chances of making the Playoffs, whereas 0.6*0.6 = 0.36 would be their chances with a 60% winrate per game... so doubling their winrate per game quadruples their Playoffs chances.

A bit pedantic, I know, and obviously in isolation for any theoretical game, it's perfectly linear.

Love the site, am working my way through the articles.
Steven.

Tangotiger had a link to the paper last week if an...

2009-10-28T15:35:33.493-04:00

Tangotiger had a link to the paper last week if anyone wants it.

One thing that struck me was how much more productive passing was than rushing using their expected points style measure. But I think that by excluding qb rushes they may have missed part of the gain of rushes. I think QB rushes may be more productive than RB rushes because they tend to be either Qb sneaks on 3rd and short or plays where the QB drops back but the defense have covered all recievers but allowig the QB to run.

They foucsed on the game theory aspect of play calling by coaches but this continues at the level of the player as well.

Also their graph didn't include inside both 10 yard lines which romer showed were very different.

Stan-I think you misunderstand. Those consideratio...

2009-10-12T19:36:07.599-04:00

Stan-I think you misunderstand. Those considerations are largely accounted for by the study. They value each play by the score advantage the result gives the offense. A 3 yd gain on 3rd and 2 is valued higher than a 3 yd gain on 2nd and 10.

In the study's model, consistency as you describe would be valued appropriately.

The biggest problem here is the foolish assumption...

2009-10-12T19:03:20.975-04:00

The biggest problem here is the foolish assumption that the offense's goal is to maximize the yardage on a particular play and that the defensive objective is to minimize the yardage on that play. It's not, not even close. The offensive objective is to move the chains. The defensive objective is to get the ball back without giving up a score. The interaction of those objectives profoundly changes the dynamics of play-calling.

On offense, I'd rather have an offense that got 4 yards every play than one that averaged 10 yards a play. Consistent efficiency on offense has a great deal more value in winning a game than inconsistent big plays. Similarly, defenses would rather create one big negative play or penalty than worry about miminizing the gain on any particular play.

In other words, it ain't about the individual play. It's about the entire possession.

The authors find that there is “substantial ne...

2009-10-06T20:01:39.558-04:00
The authors find that there is “substantial negative correlation"...

Just a thought:

Looking at offensive play selection by itself neglects the influence that defensive calls have on offensive calls.

E.g.: Coaches call "run", QB walks to the line, sees eight in the box, audibles to "pass".

In principle the initial O run-pass choices could be made by a random number generator, but if the D calls aren't random for some reason (good or bad) the final O-play choices won't be either.

FWIW.

Like Guy said: is it not likely that after a 20 ya...

2009-10-05T18:31:05.899-04:00
Like Guy said: is it not likely that after a 20 yard running gain, the best RB is tired and you have to do something else? I've seen them head to the sidelines after a play like that. And the guy who has to sprint 40 yards down the field on a passing play even if he never gets the ball ... well, you don't want him to do that twice in a row, do you?

What journal did they publish in? You might consi...

2009-10-05T13:09:01.296-04:00
What journal did they publish in? You might consider sending a condensed version of this to the letters section of that journal. It's possible the academics aren't reading this blog, though they should, and you might find some collaborators.

Whenever someone writes about play-calling in the ...

2009-10-05T11:27:13.908-04:00
Whenever someone writes about play-calling in the N.F.L., I sometimes think about the conversation Tricky Dick had with Henry Kissinger. It was Nixon, if you remember, who once sent then-Redskins coach George Allen some plays for his squad to use that Sunday.

"I think it's fascinating," Nixon reportedly told Kissinger, then the secretary of state. "It's quite a science."

To which old Henry replied, "What's the mystery? It's either run or pass."

John M. Sweeney

It seems logical to me that playcallers would alte...

2009-10-05T10:24:15.873-04:00
It seems logical to me that playcallers would alternate more after unsuccessful plays, even if it's not necessarily maximizing a team's advantage.

Consider a 2nd and 10, where a team decides to run the ball. If a run is successful, the RB might gain a first down, giving the team more flexibility to run the ball again. If he's stopped for a loss, they're put in a 3rd-and-long "passing situation." An incomplete pass on 1st down, too, will result in a 3rd and long if the 2nd down play is also unsuccessful, so it seems that a run on 2nd down would make 3rd down more managable. No playcaller wants to go 3 and out and not gain any yards.

On the other hand, a successful play might highlight certain weaknesses is a defense, whether it's a hole in the line or a gap in coverage, an exploitable secondary, etc, that could be taken advantage of. In other words, if something is working, don't change it until the defense adjusts.

Not saying any of this is logical from a statistical standpoint, just trying to understand what goes through a coach's mind when calling plays.

Brian: Very nice analysis. I'll be intereste...

2009-10-05T09:39:41.191-04:00
Brian: Very nice analysis. I'll be interested in seeing your study when it's completed.

Question on the play predictability issue: is it possible that running backs and/or receivers perform slightly less well when asked to run two or more successive plays? That is, could there be a small improvement in performance if you give your best backs and receivers a rest sometimes, even if minimax calls for you to give them the ball for a 3rd or 4th consecutive play?

Keep up the great work......