Comments on Advanced NFL Stats: (Much) More on 1st and 10 Run-Pass Balance

I think you really, really need to plot the percen...

2009-12-01T19:52:53.849-05:00

I think you really, really need to plot the percentage difference between run/pass rather than the absolutes. You can't really make out anything for the cases where there's a huge change over the field (TD% and FG%).

(Error bars would be nice too, but that's just me dreaming.)

Great stuff, Brian. Makes a very strong case. ...

2009-12-01T10:31:42.748-05:00

Great stuff, Brian. Makes a very strong case.

Do you think it would be worth looking at the pass/run EP gap at the individual team level? Or perhaps by groups of teams, based on size of the gap? Might be interesting.

Thanks Matt. I forgot the first article had the E...

2009-11-30T14:40:52.171-05:00

Thanks Matt. I forgot the first article had the EP graph. Although it looks like it follows my example from these graphs - running versus passing preference becomes indifferent and switches to favoring the run closer to the endzone.

Alex, I believe the 2% turnover increase is alrea...

2009-11-30T13:46:52.907-05:00

Alex, I believe the 2% turnover increase is already built into that EP calculation... although I do see what you are saying. It seems as though the NFL coaches are risk-averse and may even be aware that on average they should pass (or go for it on 4th down), but choose to minimize potential damage... ie coaching for the friendliest loss.

It makes sense though, if you remember the patriots-colts debacle of a few weeks ago. Many people would rather not risk winning or losing the game on 1 play, even though by the numbers it is correct... they would rather have multiple opportunities on defense to come up with a big play. Coaches/analysts often seem to have this risk-averse mentality. And of course analysts will mock you for whatever decision you made if you loose and applaud if you win, regardless of whether it was statistically correct.

Hi Brian - Just a suggestion. I think all the grap...

2009-11-30T10:39:22.849-05:00

Hi Brian - Just a suggestion. I think all the graphs, excepting maybe the average yards gained, would be more readable if you graphed the pass-run difference instead of the two values. This would get rid of your scaling issue.

Also, would you be able to make a graph of average or expected points for run versus pass across different field positions? Just eyeballing values from your graph, it looks like passing on first down from the 21-30 is worth 4.2ish points while running is worth 4.1ish (multiplying the probability of a FG or a TD by 3 or 7 and adding them together). Maybe going conservative is worth the small difference in points, given that you're 2% less likely to turn the ball over?

I assume this is your normal 8-9 years worth of da...

2009-11-30T10:39:01.125-05:00

I assume this is your normal 8-9 years worth of data. I have heard a lot of comments by the 'analysts' that the NFL has changed to a passing league lately. I've seen some evidence showing there hasn't really been much of a change... but I bet these analysts would argue that passing is now more valuable over the last few years and is a more effective strategy partially because the defenses haven't adjusted.

Is this data showing that the pass is more beneficial on 1rst down consistent over the data set, or it is increasing over time, meaning that once running was more beneficial, and now passing is the better option?

Yes, there's plenty of bias in both directions...

2009-11-30T09:35:42.256-05:00

Yes, there's plenty of bias in both directions. Teams good at passing will pass more often, making passing look better. But teams good at running will run more often too, making running look better. I wouldn't rely on the pure exact numbers here, but the fact that passing is superior to running, and not the other way around, tells us a lot.

Teh study only contains data Points from the 1st a...

2009-11-30T09:08:55.329-05:00

Teh study only contains data Points from the 1st and 3rd quarter in close games, so these effects should not be existing in those data points.

PS: The more interesting question, is, if the study contains Selection bias due to the fact that the better offenses normally pass more often.

Quick comment on the last graph : Data might be bi...

2009-11-30T08:44:52.114-05:00

Quick comment on the last graph : Data might be biased by teams trailing, and going in pass-heavy mode.

These teams :

A - are sometimes going up against softer prevent defenses, boosting their success level
B - more likely to go for it on 4th down, so less likely to punt