VFL-82-JP
Bleedin' Orange...
- Joined
- Jan 17, 2015
- Messages
- 19,653
- Likes
- 52,096
Listening to that fella, what's his name, Josh Pate or something like that, on the 247 Youtube channel Late Kick Cuts or something (sorry if I'm messing these names all up)...
He said a couple of things about computer modelling for football analysis that made lightbulbs go on for me.
One thing he said was, a turnover is worth about 4 to 4.5 points for the gaining team. I didn't realize that, but it makes sense. And it makes the General's "play for and make the breaks, and when one comes your way, score" maxim pop. Turnovers are the biggest breaks of all. And they're worth 4 to 4 1/2 points, wow. Wow.
Thinking about it some more, it makes perfect sense. Sure, you don't always score from a turnover, only sometimes. And maybe then, it's just a field goal. But you're ALSO stopping the other team from maybe scoring, as well. So that 4 - 4.5 point swing is perfectly sensible.
Turnovers are huge. No wonder CJP is pushing so hard for the big plays, the splash plays, the turnovers.
I mean, I knew all this before. Kind of. Sort of. It just never really popped into my consciousness just how many points are tied to turnovers.
Okay, that was the first eye-opener he gave me. The second was, you can't model for turnovers. Not with any fidelity at all. Because they're so unpredictable.
Even if you include in your model something like the to-date turnover advantage/disadvantage of each team playing, that is still just a very rough and inadequate piece of your predictive model.
Almost everything else can be modeled well. Maybe not ref bias, but everything else. Surprisingly accurate, too.
But not turnovers.
So if I predict your team is going to win by 6 points, but you turn the ball over twice and your opponent doesn't turn it over at all, everything else being equal, you're going to lose by 2-3 points.
The turnovers are a big part of why no Vegas bookie gets it right more than about 75% of the time. They're why DAJ (our resident math guru) does about the same in the pick-em contests. They're why ESPN's FPI does about the same.
No one can predict the turnovers, and they're huge in point value.
~ ~ ~
Okay, I see you all looking at me like I just discovered water is wet. Okay, okay, I'll be quiet. Heh, for some reason, these light bulbs just came on for me. Thank you Josh Pate, or whatever your name is.
Go Vols!
He said a couple of things about computer modelling for football analysis that made lightbulbs go on for me.
One thing he said was, a turnover is worth about 4 to 4.5 points for the gaining team. I didn't realize that, but it makes sense. And it makes the General's "play for and make the breaks, and when one comes your way, score" maxim pop. Turnovers are the biggest breaks of all. And they're worth 4 to 4 1/2 points, wow. Wow.
Thinking about it some more, it makes perfect sense. Sure, you don't always score from a turnover, only sometimes. And maybe then, it's just a field goal. But you're ALSO stopping the other team from maybe scoring, as well. So that 4 - 4.5 point swing is perfectly sensible.
Turnovers are huge. No wonder CJP is pushing so hard for the big plays, the splash plays, the turnovers.
I mean, I knew all this before. Kind of. Sort of. It just never really popped into my consciousness just how many points are tied to turnovers.
Okay, that was the first eye-opener he gave me. The second was, you can't model for turnovers. Not with any fidelity at all. Because they're so unpredictable.
Even if you include in your model something like the to-date turnover advantage/disadvantage of each team playing, that is still just a very rough and inadequate piece of your predictive model.
Almost everything else can be modeled well. Maybe not ref bias, but everything else. Surprisingly accurate, too.
But not turnovers.
So if I predict your team is going to win by 6 points, but you turn the ball over twice and your opponent doesn't turn it over at all, everything else being equal, you're going to lose by 2-3 points.
The turnovers are a big part of why no Vegas bookie gets it right more than about 75% of the time. They're why DAJ (our resident math guru) does about the same in the pick-em contests. They're why ESPN's FPI does about the same.
No one can predict the turnovers, and they're huge in point value.
~ ~ ~
Okay, I see you all looking at me like I just discovered water is wet. Okay, okay, I'll be quiet. Heh, for some reason, these light bulbs just came on for me. Thank you Josh Pate, or whatever your name is.
Go Vols!