A peek into how just about every analytics ranking system is biased by preseason ratings

Sign up, and you'll be able to customize your font size and more! Sign up

Report problem with this ad

Sep 16, 2024
7:40:38am

Indy Coug Loquacious Lummox

A peek into how just about every analytics ranking system is biased by preseason ratings

Sagarin, FPI, SP+, etc. are all using preseason weightings early in the year.

Kenpom does the same thing as well.

FEI ratings updated through Week 3: https://t.co/wUM0PEDzdp
82% preseason projection weight for teams that have played only one FBS opponent, 67% for teams that have played two, 54% for teams that have played three, 43% for teams that have played four pic.twitter.com/EErKHN226M
— Brian Fremeau (@bcfremeau) September 16, 2024

https://twitter.com/bcfremeau/status/1835658728160288774

This message has been modified

Originally posted on Sep 16, 2024 at 7:40:38am

Message modified by Indy Coug on Sep 16, 2024 at 7:41:16am

Message modified by Indy Coug on Sep 16, 2024 at 7:46:25am

Report problem with this ad

Indy Coug

Bio page: Indy Coug
Joined: Dec 20, 2001
Last login: Sep 18, 2024
Total posts: 187,725 (10,022 FO)

Report problem with this ad

Messages

Author

Time

A peek into how just about every analytics ranking system is biased by preseason ratings

Indy Coug Loquacious Lummox

Sep 16, 7:40am

Notre Dame at 8.....wow.

WJCougar

Sep 16, 7:42am

Should be a real challenge this weekend. Go Cougs!

inertiaman

Sep 16, 7:47am

Which of course, makes sense. A ranking system based entirely on 1 or 2 or even 3 games would be completely worthless.

TCuz

Sep 16, 7:52am

It would be limited, but this model is *worse* than worthless, because it creates capricious bias based on something

mvtoro

Sep 16, 10:13am

Dumb

RGGeemer

Sep 16, 7:59am

The weaker the signal, the more weight you put on your prior. That isn't bias.

Coop

Sep 16, 8:48am

I'm no numbers guy like you but I have been saying this for years.

CougaRR4L

Sep 16, 9:33am

Florida State anyone? WA beaten by WA St yet #15.

ImuaKahuku

Sep 16, 10:38am

rankings would be a laughing stock if they didn't do this however

Hill4Heisman

Sep 16, 10:47am

It would be better for them to look meaningless until they have actual meaning.

mvtoro

Sep 16, 11:14am

Or ideally, they could figure out how much to weight preseason priors, using historical data to get the best predictions

krindorr

Sep 16, 11:18am

Statistically/analytically this makes sense though; historically a better predictor than just overreacting to game or 2

krindorr

Sep 16, 11:13am

What makes the most sense is to not rank teams until at least part way in.

ImuaKahuku

Sep 16, 11:57am

Report problem with this ad

Posting on CougarBoard

In order to post, you will need to either sign up or log in.

Report problem with this ad