WTF CFB math

[pokerjoe]

So, Braxton Miller is out for O St. What's the impact (in points, duh). Looks like the line has adjusted by 5 points (17' to 12'; "points" is not the best way to express such changes, but this is a forum, so, yes it is).

 

http://www.usatoday.com/story/sports...rios/14279907/

 

There, they make the predicted impact to be 9 points. That's WTF to me. From Ohio St expecting 44 ppg down to 35 ppg?

 

Even the line move, I think, is an overreaction; this model's projection is ridiculous. But if it wasn't ridiculous, it wouldn't be public, would it? It's makers would still be hammering bets out. So it's something of a law: any ratings you find on the internet must be failed or you wouldn't be finding them on the internet.

 

Even Pomeroy's CBB ratings I believe (I wish I could remember better; it was an interesting phenomenon, the opening of his site), were already failing to win when he went public (he has mentioned, I believe, that he has found sports betting unbeatable). His are the best publicly available ratings for any sport I've ever heard of. Most are Sagarin-style near worthlessness.

 

I'm in the midst of my CFB prep, so I haven't made my own numbers yet on the Miller injury, but, generally, 3 points is a lot for a starting QB, 5 is a mega lot, 9 isn't going to happen.

 

These guys rate Winston's impact at 12 ppg, btw. Their "math" does.

 

And this is why fat guys wearing flip flops and smoking cigars can beat the Cantor suits in this biz.

 

I personally do not think it's possible to well model player impact in CFB. The samples are always really small (you would be very lucky if the injured starter and his replacement both even have 10 games on record) and team play/coaching adjustments are really complex and filled with dependencies (it can be hard to admit that RB stats are so OL dependent that starter-sub stats can be almost meaningless, especially as usage style can be so important, and game plan changes can impact stats), and the game lends itself to high emotions (so that a team can play better, rallying when a key starter is out).

 

I don't think math is the answer here, I think quantified instinct still works best. I mean guys who follow the sport and have seen such situations many times before and have tested and sharpened their instincts in a running battle with the market, sitting down and feeling out how much they want to rate the Braxton Miller injury. Putting a number on it--that's why I said quantified instinct--but trusting their judgment more than anything a program can spit out.

 

This is for football, mind you. For a sport like MLB, I'll take a good program anytime. Basketball is somewhere between the two sports.

[maron]

Margin of victory is probably less important this year than in past due to computers not having direct input into the process. The 13 who decide this can probably look past points just to the dominance of the team. Hence with OSU's cupcake schedule then it reasons the PPG might drop to the 35 or even lower