NET Rankings Release

freakboy · Post by **freakboy** » December 7th, 2022, 7:29 pm

They have mentioned that they may change the algorithm over time, so it’s impossible to know. I was hoping to backward engineer it, but if it changes Over time, that complicates things.

StanfordAggie · Post by **StanfordAggie** » December 7th, 2022, 8:53 pm

flying_scotsman2.0 wrote: ↑
December 4th, 2022, 10:35 pm

Aggie84025 wrote: ↑
December 4th, 2022, 9:59 pm
How close do ken pom rankings mimic net rankings?
They’re usually pretty close, but sometimes there are teams that are significantly different.

I heard somewhere that Ken Pomeroy was consulted when the NCAA was developing the NET. Not sure if that’s true or not.

At one point Pomeroy's offensive and defensive efficiencies were components in the ranking. I think that may no longer be the case, though.

StanfordAggie · Post by **StanfordAggie** » December 7th, 2022, 8:59 pm

freakboy wrote: ↑
December 7th, 2022, 7:29 pm
They have mentioned that they may change the algorithm over time, so it’s impossible to know. I was hoping to backward engineer it, but if it changes Over time, that complicates things.

They changed it prior to the 2020-2021 season:

https://www.ncaa.com/news/basketball-me ... et-2020-21

I would think that to successfully reverse engineer the algorithm, one would probably need to download each daily release of the NET rankings and try to fit a model that works consistently for each day. End-of-year data won't give a large enough sample size, particularly if they keep tweaking the formula every year or two.

freakboy · Post by **freakboy** » December 7th, 2022, 9:00 pm

StanfordAggie wrote: ↑
December 7th, 2022, 8:59 pm

freakboy wrote: ↑
December 7th, 2022, 7:29 pm
They have mentioned that they may change the algorithm over time, so it’s impossible to know. I was hoping to backward engineer it, but if it changes Over time, that complicates things.
They changed it prior to the 2020-2021 season:

https://www.ncaa.com/news/basketball-me ... et-2020-21

I would think that to successfully reverse engineer the algorithm, one would probably need to download each daily release of the NET rankings and try to fit a model that works consistently for each day. End-of-year data won't give a large enough sample size, particularly if they keep tweaking the formula every year or two.

Yes, that was exactly my plan. I spoke to a few people like Ken Pomeroy to try to get some idea of the type of model Google might have been using.

flying_scotsman2.0 · Post by **flying_scotsman2.0** » December 7th, 2022, 10:11 pm

Except the NCAA can change the NET algorithm at any time, right? So who knows how many mid-season changes they make...

Post by **ProvoAggie** » December 7th, 2022, 10:19 pm

flying_scotsman2.0 wrote: ↑
December 7th, 2022, 10:11 pm
Except the NCAA can change the NET algorithm at any time, right? So who knows how many mid-season changes they make...

Correct. It's hard to know if they have. We know they made the major changes once but that doesn't mean that they haven't done any minor fine tuning throughout a season.

Upcoming Events

Event Results