8
Started By
Message
Week 5 Power Ratings and SEC predictions
Posted on 9/27/26 at 10:53 am
Posted on 9/27/26 at 10:53 am
This is not a ranking, though many of you will not understand that. This is simply a measure of who would be favored over who on a neutral field - that's it. It doesn't consider merit or who deserves what.
An example for the tards: Notre Dame (28.2) would be a 6 point favorite over Florida (22.2) on a neutral field, because 28.2 - 22.2 = 6.
An example for the tards: Notre Dame (28.2) would be a 6 point favorite over Florida (22.2) on a neutral field, because 28.2 - 22.2 = 6.
Posted on 9/27/26 at 11:07 am to MilkJug
Western Michigan beating Penn State is going have a pretty significant boost to Notre Dame's numbers... for now, at least. If Penn State tanks then the win won't be as impressive.
Posted on 9/27/26 at 11:19 am to skrayper
quote:
Western Michigan beating Penn State
Either you meant Wisconsin, or the joke went over my head.
But yes, you are correct - that is how the model works.
Posted on 9/27/26 at 11:48 am to MilkJug
I do not envision Notre Dame being a neutral field favorite against Georgia, Texas, Ohio State, Bama, Florida, Oregon, or LSU.
Posted on 9/27/26 at 11:49 am to MilkJug
Eight teams from 2-10 would easily beat ND. I’d say that ND vs. Utah would be a toss up.
Posted on 9/27/26 at 11:50 am to bigbopper
quote:
I do not envision Notre Dame being a neutral field favorite against Georgia, Texas, Ohio State, Bama, Florida, Oregon, or LSU.
Let them. Easy $$ imo
EDIT: remove Oregon. They are terrible
This post was edited on 9/27/26 at 11:51 am
Posted on 9/27/26 at 12:20 pm to MachoMan
Oregon didn't really skip a beat when Raiola went in. Not garbage.
Posted on 9/27/26 at 1:09 pm to MilkJug
What is the source?
Posted on 9/27/26 at 1:13 pm to MilkJug
what is this nonsense based on?
Posted on 9/27/26 at 3:08 pm to narddogg81
quote:
what is this nonsense based on?
Bayesian ridge regression with a Kalman filter.
It's only goal is to predict the spread first, then the total. Football scores are extremely noisy, which is why they are hard to predict. When Vegas releases their spreads, they're based on a similar model to mine. Then Vegas lets the market take over to push or pull the spread.
My model (over a season) hits and misses at pretty much the same clip as Vegas.
This post was edited on 9/27/26 at 3:09 pm
Posted on 9/27/26 at 3:11 pm to MilkJug
Northwestern making these list means your entire system is broken. No need to explain to us "tards".
Posted on 9/27/26 at 3:17 pm to MilkJug
Can you post your ratings from last year at the end of the regular season?
Posted on 9/27/26 at 3:18 pm to MilkJug
quote:
This is simply a measure of who would be favored over who on a neutral field - that's it. It doesn't consider merit or who deserves what.
Based on what? You have to be applying some kind of data weighting and analysis, or is this just your feels?
Posted on 9/27/26 at 3:25 pm to GreenieTiger
quote:
Northwestern making these list means your entire system is broken. No need to explain to us "tards".
Most teams have only played 1, maybe two good teams. And if they do good against those good teams, the model is going to be high on them - simply because of the lack of data. Northwestern blew out Colorado and almost beat Indiana. The model's ratings change week to week as it gathers more data.
But, yes, you are a tard and I don't expect you to understand how any of this works.
Posted on 9/27/26 at 3:27 pm to Sun God
quote:
Can you post your ratings from last year at the end of the regular season?
I don't keep weekly snapshots of ratings. I have the ratings from the day after the national championship last season.
Posted on 9/27/26 at 3:29 pm to MilkJug
Interesting
Posted on 9/27/26 at 4:15 pm to MilkJug
I understand how your model works secRant genius. I am telling you that your MilkJug model is a piece of shite.
Posted on 9/27/26 at 4:16 pm to Geauxgurt
quote:
Based on what? You have to be applying some kind of data weighting and analysis, or is this just your feels?
I'm not weighting anything by hand.
Every FBS (games against FCS opponents are ignored) game this season goes into one big system of equations. The unknowns are each team's offense and defense rating, plus league-average scoring and home-field advantage:
Home score = league avg + home offense - away defense + home advantage (~3)
Away score = league avg + away offense - home defense
Ridge regression solves all of it at once. A team's rating is fit against the specific opponents that team played. What the model is "trying" to do is achieve the lowest MAE (Mean Absolute Error) possible - which is the whole point of the model - the outcome is the ratings/score predictions.
A Kalman filter then updates the ratings week to week, so a blowout 4 weeks ago counts for less than the most recent game a team played - the reason for this is that teams get better (and some worse) throughout the season. It's about a 5% decay each week. But this isn't intuitively "weighted", that "5%" is not a number I chose. The Kalman filter adjusts this decay itself accordingly to reach it's lowest MAE possible.
How does it know it's current MAE? It tests itself against previous games.
So it's not a bunch of dials that I decided matters. It's literally all just Math - and admittedly, football is more than just Math - there are a lot of unknowns and noise that effect the outcomes of games - which is why we like it so much.
This post was edited on 9/27/26 at 4:20 pm
Posted on 9/27/26 at 4:16 pm to GreenieTiger
quote:
I understand how your model works
You don't
Page 1 of 1
Popular
Back to top
Follow SECRant for SEC Football News