APBRmetrics

This is just a starter post on the topic.

I recently viewed the 2012 similarity scores produced by Schoene to project draft picks and the listed closest comparisons seemed weak overall imo from the view of today at least and maybe from 2012 as well. (can be found at basketball prospectus. Link will be posted later. On phone, can't do now.)

I'll probably have more comments later. The topic hopefully will review and compare Schoene with other similarity system out there now.

But for now, my main observations are:

It might be better to break player comparisons into offensive and defensive similar rather than finding guys who are close across both.

On offense the comparisons should probably be plus or minus 25-35% on usage. 50 plus percent differences in usage seem like they should be disqualifying to me, unless folks can cite many cases where there were such discrepancies in college stats and yet there was a good NBA similarity score later. Maybe there are decent standard usage adjustments based on other NBA quality players on the roster and strength of conference / schedule that can be made to put guys into tighter NBA usage range groups from which to focus the similarity lists. (I know Schoene has other usage adjustments.)

It is probably better to list the 3-5 best similarities rather than just one.

Are you looking for similar players in the league today? Past players similar to current ones?

Most commonly similarity systems are used to find NBA players (current or recent) who college players about to be drafted were similar to in college. There would be good uses for a current NBA player to current or recent NBA player similarity database too though- how much to pay young players, who they might best fit with, etc. I left open the possibility to discuss both, but expected there would be more to discuss on the first usage.

I don't know that I would say most commonly; SCHOENE for the NBA long precedes my using the same similarity formula for college players. Because of the small pool of players, especially at extreme ages, I tend to view college similarity as more of a novelty than a primary part of the analysis.

It is probably better to list the 3-5 best similarities rather than just one.

That is definitely better! It's no coincidence that I was working for a team when I posted the closest match alone.