PDA

View Full Version : Most important individual stat in NCAA football?



smokymountainrain
11-20-2008, 12:15 PM
Pass Efficiency Rating

Actually, I have no clue how significant this is, but I found it interesting nonetheless that the teams of top 12 QBs in the country have a combined record of 111-14. Or an average record of 9-1.

Jeff Lebowski
11-20-2008, 12:16 PM
We need Indy to show up and fill us in on the PE differential. He claims that is the key stat.

All-American
11-20-2008, 01:42 PM
Well, I'm no Indy, but . . .

BYU Pass Efficiency: 166.84
BYU Pass Efficiency (Defense): 116.46
PE Differential: 50.38

Utah Pass Efficiency: 144.42
Utah Pass Efficiency (Defense) 113.46
PE Differential: 30.96

PED Differential: BYU +19.42.

All-American
11-20-2008, 01:44 PM
And Indy on why that matters:

http://www.cougarboard.com/nologin/message.html?id=2279662

And his breakdown:

PED Diff W L Win% Avg MOV
125 to 150 3 0 100.0% 47.3
100 to 125 36 0 100.0% 38.6
75 to 100 165 4 97.6% 30.9
50 to 75 477 43 91.7% 21.3
25 to 50 829 213 79.6% 13.1
0 to 25 934 589 61.3% 4.3

RockyBalboa
11-20-2008, 02:02 PM
Well, I'm no Indy, but . . .

BYU Pass Efficiency: 166.84
BYU Pass Efficiency (Defense): 116.46
PE Differential: 50.38

Utah Pass Efficiency: 144.42
Utah Pass Efficiency (Defense) 113.46
PE Differential: 30.96

PED Differential: BYU +19.42.

Help an idiot out....so this is good for BYU right?

pelagius
11-20-2008, 02:05 PM
Help an idiot out....so this is good for BYU right?

Yes, Indy's prediction would be something like BYU should win by 4-6 points (of course you should probably throw in a home field advantage adjustment at the very least which would reduce it to like 1-3 points).

* just based on the table its 4 points but I assume Indy would want to adjust it upward more because 19.4 is towards the upper end of the lowest differential breakpoint.

ERCougar
11-20-2008, 02:06 PM
We need Indy to show up and fill us in on the PE differential. He claims that is the key stat.

I think I have to go with points being the key stat. That or PD (point differential).

pelagius
11-20-2008, 02:14 PM
I think I have to go with points being the key stat. That or PD (point differential).

No doubt PD (point differential) is a great stat but it has its problems as an ex ante predictor.

ERCougar
11-20-2008, 02:26 PM
No doubt PD (point differential) is a great stat but it has its problems as an ex ante predictor.

I wonder how PED compares to PD (ex post=over past games?) as a predictor.

Hot Lunch
11-20-2008, 02:49 PM
What about turnovers?

pelagius
11-20-2008, 03:00 PM
I wonder how PED compares to PD (ex post=over past games?) as a predictor.

You're right you can use PD in an ex ante way.

That's what the computer models are really doing to some degree. You use the information in past game score (margin of victory) to rank teams and predict. The MOV models are not as good of predictors as the Vegas spread. So just go with spread as your ex ante predictor. Indy claims he can beat the spread about 60% of the time with a modified version of PED (I am skeptical of this claim, I believe it worked this well for some sample period, but I am skeptical it will work that wll going forward, but if it does Indy should be become a very rich man).

All-American
11-20-2008, 03:25 PM
You're right you can use PD in an ex ante way.

That's what the computer models are really doing to some degree. You use the information in past game score (margin of victory) to rank teams and predict. The MOV models are not as good of predictors as the Vegas spread. So just go with spread as your ex ante predictor. Indy claims he can beat the spread about 60% of the time with a modified version of PED (I am skeptical of this claim, I believe it worked this well for some sample period, but I am skeptical it will work that wll going forward, but if it does Indy should be become a very rich man).

The problem is that PED is only good towards the end of the season, once you've gotten enough data to be able to figure out how good the two teams are. It's good for November games and for bowl games. Maybe.

All-American
11-20-2008, 03:26 PM
What about turnovers?

Somewhat accounted for in PED, as interceptions figure into the stat.

pelagius
11-20-2008, 03:29 PM
The problem is that PED is only good towards the end of the season, once you've gotten enough data to be able to figure out how good the two teams are. It's good for November games and for bowl games. Maybe.

This would be true for a MOV based predictor as well. But, yes, that problem does limit Indy's ability to potentially make money off that strategy.