I still havenāt heard a good explanation from Connelly, KenPom, Torvik, etc. on how exactly the preseason influence is actually removed. There is a good amount of relativity in the data, especially as it relates to SOS. I just donāt see how you can actually remove the starting point and its influence on downstream data/analytics.
My understanding is that itās diluted out as games go on.
Like starting with one drop of apple juice (preseason rank) then each new game adds an ounce of water. By the time you get to a 12oz or 16oz glass you arenāt tasting that apple juice any more.
1 Like
Yeah but % and timing of dilution is relevant (and a human choice, as are what Connelly has decided makes a good football team). And therefore it kinda matters when you have your best performances and when you play quality opponents (aka itās better for those to happen later in the season for a team like UVA whose preseason rating was crap). Thatās not good data, imo, especially for a sport with a much smaller amount of data generally than other sports.
The opponent rankings, at least in KenPom, are as of the end of season, not as of when you play them.
So everything adjusts until all games are done. So last season our Kenpom ranking kept changing until the season was over, even after we were done playing. Make sense?
1 Like
We have a great chance to go undefeated at home this year. Anything less than an undefeated season at Scott at this point will be a pretty massive failure in my eyes. Thereās no better way to resurrect this fanbase (in terms of home attendance) than just flat out not losing games there.
7 Likes
Hereās a post from kenpom on how preseason ratings are gradually removed: https://kenpom.com/blog/what-happens-to-preseason-ratings-when-its-not-preseason-anymore/
I think there will always be an indirect influence of preseason ratings bc every teamās rating depends on every other teamās rating and those interconnections were established when preseason priors were active but by late season, the weight of accumulated game data vastly outweighs the starting assumptions so that the ratings become primarily empirical
2 Likes
Hmmā¦Iām not 100% sure on the mathematics, but my intuition is that these models essentially get recomputed as new data comes in, so the value of those early season interconnections also gets updated as the preseason prior gets faded out.
2 Likes
Maybe, I donāt think the preseason issue is a big deal tbh. If a team is good theyāll rise quickly (or vice versa).
1 Like
But isnāt that the same thing as basically overweighting the games that happen after preseason adjustment rolls off? And how does SOS work, especially since the overwhelming majority of noncon games happen when preseason ratings are most heavily weighted?
(Apologies to all for me making the same point on two separate threads)
I believe everything is constantly recomputed. So as the season goes on, those early season games will also be re-evaluated as if there were no preseason weighting. Iāve never played with them, but I believe the tools on Torvikās site let you do this anytime you want during the season.
2 Likes
Oops, I see AQ already said it better than me in the other thread
Based on the transitive properties of college football wins, Virginia is better than Miami.
https://twitter.com/cfbonfox/status/1979373406689882575?s=46&t=vkjgQUekzGC7z44tIfnIRQ
4 Likes
This is a great stat. WSU also has the 2nd most missed tackles in the country, and theyāre well aware of itā¦
https://twitter.com/gregwwoods/status/1977863240752181559?s=46&t=WJO4FuELyYld9cQ4Y92CZw
They either make a big improvement tomorrow or we rush for 300 yards.
6 Likes
Whatās the best way to watch the CW tonight? I have the espn bundle.
You need a streaming plan that has a live local TV add on option like Hulu or YouTube.
Or a digital antenna.
1 Like
Go to your local Walmart, target etc. and get a digital antenna. Under $30, takes maybe 10 minutes for the initial scan. Iāve had no problems with it it just isnāt the full 4K offered.
5 Likes
Wazzu has 3 losses against teams that are 16-2 and beat SDSU which is 5-0 against everyone else.
They are a pretty good team.
5 Likes
Itās a strange morning when I applaud Kirk Herbstreit.
2 Likes
Eh Stanfordās lost to teams that are a combined 20-5 (or 16-3 if you exclude SMU last week).
2 Likes