APBRmetrics

A bit embarrassingly, I realized today that though I used per 40 minute pace adjusted numbers in running my regression, when I pulled the data for the 2014 class, I grabbed per game numbers. As you might guess, this significantly skewed the results. Here are the updated results, sorted by predicted 3rd year xRAPM:

edit: image deleted because results were wrong.

Interesting that Alec Brown rates well on your list, I had him as top 25 talent. Pretty interesting combo of outside shooting/spacing, center length and mobility.

nbacouchside wrote:A bit embarrassingly, I realized today that though I used per 40 minute pace adjusted numbers in running my regression, when I pulled the data for the 2014 class, I grabbed per game numbers. As you might guess, this significantly skewed the results. Here are the updated results, sorted by predicted 3rd year xRAPM:

Curious where Javon McCrea would rank - he was easily the highest rated non draftee in my model - and about every other model I've seen.

Also looking at it - I am shocked how low Shabazz Napier is ranked in the 1 year.

Ugh. So I transposed blocks and steals in my calculations for the 2014 class, because I'm a spreadsheet dyslexic apparently. Fixed that problem and also found a formula that does a little better predicting rookie xRAPM OOS. Updated results (including Javon Mcrea for Statman):



I also ran the year 3 projections through a super simple aging curve to look at age 28 productivity, top 10 there:

VJL, do you post at CH? I just wanted to say that I think you do excellent work.

I know the Cavs people do not like to see this, but I totally think they should do whatever they have to do, to get Kevin Love (aside from trading Lebron of course). Also Luol Deng with better defense seems like a reasonable ceiling for Wiggins. His ball handling is sorely underdeveloped.

I would love to see LeBron/Love/Wiggins on the same team though. I think those three would have better chemistry than a LeBron/Love/Kyrie trio, due to the latter's ball-dominance. Not that that's actually in play, but just a thought.

Hi all, sorry for gravebumping this thread, and sorry in advance because I don't actually have 2014 results for my draft rater, but I think there's enough interesting stuff that it's worth posting anyway. Much thanks to James Brocato for inspiring this work, and to Layne Vashro who gave me his college stats database!

This paper was written up as a submission to SSAC, which is why it's in such an official looking format, and it's possible that SSAC is mentioned at some point in the paper. To be absolutely clear, this paper was not accepted and will not appear at SSAC!

First, here are the full retrodiction results in a google doc. It only has results through the 2013 draft, and it only has results for players who played significant NBA minutes. All results are in-sample (although the sample size is large enough that by my understanding, this should not significantly skew results).

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Second, here is the paper itself. I don't think I can upload the PDF here, so it's simply a crude pile of screenshots of the PDF. I would be very happy to answer any questions about my methods and to listen to any criticism. In a lot of cases I'll probably give the lame answer that I was in a bit of a time crunch as I finished the project (the submission deadline fell right in the middle of finals period), but I'm sure there were also some real mistakes on my part!
















I don't currently have plans to continue this project or to work on basketball analytics in general, but I would be excited to talk to people who are interested in this kind of stuff. In particular, topics I'm excited about are nonlinearity corrections, methods of approximating uncertainty in player/team ratings and projections, and possible improvements to the current one-size-fits-all aging curve.

Looks like good work.

Steals 4th on first order importance (and closely followed by 5th and 6th) rather than the runaway leader found by dome other studies?

I hope you aren't shutting down your basketball analysis efforts entirely. And because of non-selection. If were to guess, I'd guess that the topic might have been considered too applied (or popular / ambitious) to be highlighted in a limited set and in this setting. Draft models might be consider too primitive right now; though this goes beyond in some ways. The writing style seemed pretty easy to follow but maybe not as formal as typical? But, again, the work itself seems solid to me. I would think it would be a good article at a analytic internet site (in addition to this one) or maybe JQAS or some student research focused outlet?

Oh no, I'm definitely not shutting down just because it got rejected (although in the unlikely event it had gotten accepted that sure would've made me think twice!). I just ran myself into the ground a little bit with my work on this paper and need to take a good long break.

Thanks for this Nathan, will try to read this tonight, and Ill post if I have any questions. Dont get discouraged!

Awesome. Glad you ended up reporting this here. Surprised the paper didn't get accepted.

Probably was going to be hard for basketball to get more than 2 of 10 this year. They probably wanted more breadth across sports, this year and from here on.

Nathan wrote:Oh no, I'm definitely not shutting down just because it got rejected (although in the unlikely event it had gotten accepted that sure would've made me think twice!). I just ran myself into the ground a little bit with my work on this paper and need to take a good long break.

Good work Nathan - I was actually thinking about your work a few days ago and what came of it. I'm glad you got it submitted, I wish it had been selected.

Maybe talk to Dr Maymin about whether he'd be interested in your piece.

http://journalofsportsanalytics.com/

Perhaps he or another associated would offer useful advice on edits or extensions.
Just a thought / guess. Maybe you could sell him on value of bringing students up to top level. A trial, that could become a path with benefits for journal as well as student and public.