Orangezoo
In the wind
- Joined
- Aug 26, 2011
- Messages
- 40,785
- Like
- 95,034
Something else to take into consideration that I've been looking at as well. With all the other non RPI metrics out there, just a quick scan shows that aside from those teams in far weaker conferences, the biggest factor in rankings seems to be total point differential. We sat around + 100 for the year, where teams like UL or even Penn St , Oklahoma etc sit higher than us on BPI, Sagarin, Ken Pom with far bigger yearly point differentials ranging from 150 plus to over 200. If those are used with any conviction for bubble analysis it is a shame. I won't have the time but a sound analysis finding the most impactful variable should find point difference carrying a far too heavy weight. Which is straight out of the old BCS playbook.
I'm sure none of them intended this but when you try to build a model with smaller components of the same larger data set, it is often going to reflect the more basic, blended result. Better efficiency is subset of how much better than your opponents you are on the scoreboard which is not an indicator of overall quality. It's who, not how.
Also funny is that despite the rankings due to point differential, the quality win/Loss records don't change.
I'm sure none of them intended this but when you try to build a model with smaller components of the same larger data set, it is often going to reflect the more basic, blended result. Better efficiency is subset of how much better than your opponents you are on the scoreboard which is not an indicator of overall quality. It's who, not how.
Also funny is that despite the rankings due to point differential, the quality win/Loss records don't change.