0
Started By
Message
Posted on 10/19/26 at 1:52 pm to Quicksilver
quote:
It'd be something like team x returns y OG this week. Compare team x's offensive EPA the first 3 weeks where y OG played vs. the two weeks he didn't and apply a weight.
The harder part is making that a reproducible prompt that, to your point, doesn't require weekly interjection but AI tools do make it a bit simpler. What makes injuries stupid to try and predict is the lack of official injury news. In the NFL, you know game participation lists are required an hour and a half before kickoff so you can schedule prompts for those slates to run. Half the time you're getting official word that college players are out literally minutes before kickoff.
Then you have situations where someone like Trammel Jones is starting over Aaron Philo. We should discount Florida some because they're missing a starting QB but how much? We have no data with Trammel at QB.
Yeah, Oklahoma might also improve with their backup QB.
I think it would require some next level AI research to accomplish. You'd have to rely on recruiting rankings for some, which isn't the best base. Consider how many years they had to get coached up, etc.
I figure there is equal amount of chance to make something worse as it is better. Maybe one day AI can do it accurately.
Posted on 10/19/26 at 1:52 pm to 3down10
quote:
I haven't done power rankings directly in over 10 years. It seems impossible to me to truly take injuries into account. I consider that to be human input, so I typically avoid that.
this is 100% correct.
Posted on 10/19/26 at 1:56 pm to Quicksilver
quote:
Then you have situations where someone like Trammel Jones is starting over Aaron Philo. We should discount Florida some because they're missing a starting QB but how much? We have no data with Trammel at QB.
This is the main issue. The problem you're trying to solve is not "what is Philo's impact to Florida's success?". The problem you're trying to solve is "what is the success or impact delta between Philo and Jones?" And that is almost an impossible question to answer.
It just so happens that Jones probably gave them better chances against Texas. Just like Chambliss was better for Ole Miss last year than Simmons.
Posted on 10/19/26 at 2:21 pm to MilkJug
What type of function do you use to weight Margin of victory? I assume some type of sigmoid function, but curious where the cutoffs are.
SP+ for example caps MOV at 50 points
SP+ for example caps MOV at 50 points
Posted on 10/19/26 at 3:09 pm to MilkJug
quote:
- Sorry, but Kentucky winning games by the skin of their teeth and with a lot of help is not impressive to a score prediction model.
Guess you have a column in your spreadsheet to account for bad calls against LSU.
Wouldn't be surprised since....
...Notre Dame's power rating is 68% higher than a team that just beat them.
SIXTY EIGHT PERCENT. That's not insignificant.
Posted on 10/19/26 at 4:08 pm to MilkJug
Stats are fun, but they are just the result of how the actual game played out.
LSU beat the brakes off of an MSU team that didn't have its starting QB or RB. If that kinda thing doesn't factor in....this is garbage.
LSU beat the brakes off of an MSU team that didn't have its starting QB or RB. If that kinda thing doesn't factor in....this is garbage.
Posted on 10/19/26 at 4:30 pm to MilkJug
That is pathetic.
Go back to tge drawing board because your model isn't mathing.
Thats the Lena Dunham of "models."
Go back to tge drawing board because your model isn't mathing.
Thats the Lena Dunham of "models."
Posted on 10/19/26 at 4:58 pm to PurpleandGeauld
quote:
LSU played without Trey'Dez Green, Whit Weeks, and 2 other starters. Princewill was out over half the game as well iirc.
States QB was a walk-on whose last start was in Mississippi private school ball. His dad did donate enough for him to get Dak's number, though.
Posted on 10/19/26 at 5:20 pm to Dixie.Reb
quote:
What type of function do you use to weight Margin of victory? I assume some type of sigmoid function, but curious where the cutoffs are.
MOV is not capped.
Posted on 10/19/26 at 5:33 pm to TDCat
quote:
Guess you have a column in your spreadsheet to account for bad calls against LSU.
This is a lot more complicated than a spreadsheet. There is zero favoritism in the model.
quote:
...Notre Dame's power rating is 68% higher than a team that just beat them.
SIXTY EIGHT PERCENT. That's not insignificant.
That's not at all how this works. The rating figure you see is adjusted by the league avg points/game. So if you want to play your ignorant, "I have no clue how this works" retard game, you should be using their current rating + the league avg points/game (24.4):
ND: 50.7
BYU: 40.1
Once you have this, you are still fricking way off. This isn't a linear scale of strength. These figures don't mean LSU (24.1) is twice as good as Missouri (12.0). The figure you see is the predicted spread between that team and the league avg team. So take the most average team in the country, ND would be a 26.3 pt favorite over that team. Make sense? didn't think so...
Sorry, but your team isn't nearly as good as you think they are. And you aren't nearly as smart as you think you are.
Posted on 10/19/26 at 5:51 pm to MilkJug
Week 8 Pretender Rankings
There, I fixed the title for you.
Stop it with these retarded posts. You are just embarrassing yourself.
There, I fixed the title for you.
Stop it with these retarded posts. You are just embarrassing yourself.
Posted on 10/19/26 at 7:01 pm to MilkJug
quote:
MOV is not capped.
Well it’s your model, but that explains my issue with LSU and ND. Guess we will find out in a month how predictive blowouts are
Posted on 10/19/26 at 7:29 pm to Dixie.Reb
quote:
but that explains my issue with LSU and ND
how?
Posted on 10/19/26 at 7:53 pm to MilkJug
ND MOV:
28
52
17
39
11
51
LSU MOV:
41
31
29
39
31
More ND than LSU, but capping impact of MOV at say 35 would reduce their strength rating Im pretty sure.
I also suspect, even without a cap the difference between a 24 and 31 point win is a lot less predictive than the difference between a 7 and a 14 point win. Which is why I asked if you had an S-curve type weighting function
28
52
17
39
11
51
LSU MOV:
41
31
29
39
31
More ND than LSU, but capping impact of MOV at say 35 would reduce their strength rating Im pretty sure.
I also suspect, even without a cap the difference between a 24 and 31 point win is a lot less predictive than the difference between a 7 and a 14 point win. Which is why I asked if you had an S-curve type weighting function
Posted on 10/19/26 at 9:33 pm to Dixie.Reb
But this is just capping MOV to cap MOV. What's the reason? You realize this would effect other teams, right? This is injecting human opinion - which is not good for predictive models.
Capping MOV is directly hurting a model that is trying to be predictive. How do you not see this?
Capping MOV is directly hurting a model that is trying to be predictive. How do you not see this?
Posted on 10/19/26 at 9:48 pm to MilkJug
quote:
Sorry, but your team isn't nearly as good as you think they are. And you aren't nearly as smart as you think you are.
Physician, heal thyself.
Posted on 10/19/26 at 9:49 pm to Dixie.Reb
Also, the SP+ you see on ESPN's website is the predictive SP+. That model does not cap margin of victory. I don't even think this model considers MOV. Connelly's Resume SP+ caps MOV - that is his "ranking" model, which is not on ESPN's website. I also have a ranking model (not done with it yet), that factors in running up the score, playing the 2nd and 3rd string players, etc. But that's a ranking, not a predictive model and not what you see in my OP.
Posted on 10/19/26 at 11:52 pm to MilkJug
Are you weighing turnovers or havoc rates?
Posted on 10/20/26 at 7:55 am to Quicksilver
quote:
Are you weighing turnovers or havoc rates?
No. And I've thought about this, but haven't looked at ways to incorporate it. But I'm almost certain it's going to inject human opinion and create noise that makes the model more inaccurate.
Page 3 of 4
Popular
Back to top
Follow SECRant for SEC Football News