#94 in KenPom

[blockm2]

From their methodology, their preseason ranking is based on prior year - and they slowly wean the model off prior year data until they no longer use it at all around the third week of January. So right now, kenpom is weighting last year's performance at >50% of their overall model, I'd think.

Actually the KenPom preseason rankings are kinda complicated. He doesn't explain that exact weights of things, but does admit they include several years worth of team level data for teams that have the same coach, they include the amount of money programs spend on basketball (since that helps predict their future success), returning players, incoming transfers and recruits, etc.

As he explains, if you wanted to predict the Pac 12 standings 15 years from now, you'd have a pretty safe bet putting schools like Arizona and UCLA at the top and schools like Oregon State and Washington State at the bottom knowing nothing more than how much money they currently spend on basketball.

It's obviously an inexact science. But every year he measures the preseason ranks against the end of season ranks to see how it is doing and assess tweaks that can be put into the formula for next year. I'd say overall the preseason rankings are useful in aggregate, but can be off for one reason or another on a particular team.


Bart Torvik of T-rank does a similar set of preseason ratings that wean down in importance as games get played, but interestingly he attacks the predictions from purely a roster standpoint projecting individual players production based on previous seasons and recruiting rankings. KenPom's preseason is based almost entirely on team level data (except for accounting for production that returns/leaves for graduation and NBA and recruiting ranks for adding elite recruits).

 

[Greene Rice FIG]

Bart has us at #59...i think the algorithims are similar with the exception that (a guess) bart is only using current year stuff

 

[Greene Rice FIG]

actually i am not completely correct

http://adamcwisports.blogspot.com/p/every-possession-counts.html?_sm_au_=i5VQ1ZfQsP4NtFtq

 

[Mr_Twister]

Does the prior season data factor in that rosters and coaching staffs change?

 

[RUChoppin]

Actually the KenPom preseason rankings are kinda complicated. He doesn't explain that exact weights of things, but does admit they include several years worth of team level data for teams that have the same coach, they include the amount of money programs spend on basketball (since that helps predict their future success), returning players, incoming transfers and recruits, etc.

Yeah, the lack of weighting in his methodology makes it hard to gauge things in the early season. It does look like the prior year data phases out by late January, though.

 

[blockm2]

actually i am not completely correct

http://adamcwisports.blogspot.com/p/every-possession-counts.html?_sm_au_=i5VQ1ZfQsP4NtFtq

To me, T-rank from Torvik is probably about 95% as accurate/useful as KenPom, but it's free to use and has far more interesting capabilities in terms of ranks and searches and tools and ways of presenting the information.

I pay for KenPom every year because it's only $20, but I probably use T-rank slightly more because you can do more with it.

 

[ScarletDave]

Up to #68 in the NET pre-fordham game. Last team in? :stuck_out_tongue_winking_eye:

 

[GoFish2]

Up to #68 in the NET pre-fordham game. Last team in? :stuck_out_tongue_winking_eye:

Dropped all the way down to 112. Damn.