Pomeroy watch

[BlueSince92]

As of 11:13 PM EST we’ve moved up 9 spots on the strength of our opener, from 43 to 34. Only half tonight’s results are tallied so far.

Will this mean much?

Obviously not since our results against Duke will determine our actual basic ranking for the rest of this calendar year for the most part. Otoh we have a week to kill and we can watch Duke’s rating getting adjusted too.

 

[ukcatz12]

Just for the record, Kenpom uses last year's data until at least the new year. It's slowly filtered out between now and then, but it won't be 100% reflective of this current season until around January 1 or so.

 

[BlueSince92]

Well in a sense it has to use last year’s data before there is any of this year’s data to go on. He has to rank somehow preseason. But this year he’s tweaked his preseason rankings to account for transfers.

So to the extent that’s weighted, our initial (43) ranking is our players’ performance last year at their respective schools. If we clobber the field as I hope we do, we could think of Ken Pomeroy’s original forecast as bonehead. Or, alternatively, we could think of the difference in forecast versus outcome as an accurate measure of Mark Pope’s ability to make a team greater than the sum of its parts.

 

[JonathanW]

Well in a sense it has to use last year’s data before there is any of this year’s data to go on. He has to rank somehow preseason. But this year he’s tweaked his preseason rankings to account for transfers.

So to the extent that’s weighted, our initial (43) ranking is our players’ performance last year at their respective schools. If we clobber the field as I hope we do, we could think of Ken Pomeroy’s original forecast as bonehead. Or, alternatively, we could think of the difference in forecast versus outcome as an accurate measure of Mark Pope’s ability to make a team greater than the sum of its parts.

You are right and wrong. The most appropriate thing to do, especially now with every team having many more new players due to transfers, and what I used to do, is not to run your model until you have enough data for it. For my model that was after about every team had 6-7 games.

 

[BlueSince92]

You are right and wrong. The most appropriate thing to do, especially now with every team having many more new players due to transfers, and what I used to do, is not to run your model until you have enough data for it. For my model that was after about every team had 6-7 games.

Mitigating early low data confidence by delaying your early predictions is a perfectly reasonable approach. Doesn’t make me wrong about anything. Would be an odd choice for Ken since he makes a lot of money on this and has competitors.