APBRmetrics

This might be interesting to some.

The Kings will seek analysis on players from the public and eventually select two people to be in their draft room when they make their pick on June 26. Essentially the Kings are looking for amateurs to offer their analytics ideas to the team for free and in return they will be granted access

The team has launched a website where would be analysts will be asked to give rankings of draft prospects and their statistical methodology. Contestants are to use information from publicly available statistical databases to create their systems.

Link to the challenge

Has anyone looked into this? This person tried to project NBA performance using NCAA stats

http://marginalrevolution.com/marginalr ... draft.html

colts18 wrote:http://marginalrevolution.com/marginalr ... draft.html

To maybe spare some people the time of downloading/reading this.. the paper uses Berri's Wins Produced to gauge a player's production

And his model is a black box

The challenge is only open to residents of the U.S.

That leaves me out. I'll be rooting for people of this forum, then

I'm out of US too though I wasn't planning to enter.

I will say, if someone has a statistical/analytical model good enough to get hired by the Kings, I don't think they need the contest. I mean the draft community is small enough that APBR, RealGM, twitter or reddit notice everything new that comes out and it even said in an article Pete D'Alessandro is an avid lurker of those sources for new ideas. Probably better off trying to make the next Shut Up And Jam and getting hired out of that, than sending your stuff to the Kings

Dr Positivity wrote:And his model is a black box

All true, although the author is not just some random person, it's Philip Maymin who's posted here a couple of times, presented at the Sloan Conference, and now co-edits the Journal of Sports Analytics.

Which doesn't mean that it's a good model, or bad ... knowing only that it's a machine learning model doesn't give us enough info to judge.

I applaud the Kings for trying this. I hope they think about the advice given and also about the process for gathering and considering the advice. There is a lot that could be learned about how to learn and make decisions in general and in particular using the input of large groups / outsiders.

colts18 wrote:Has anyone looked into this? This person tried to project NBA performance using NCAA stats

Actually, my draft model will project entire NBA careers year by year using their college stats (adjusted for pace/SoS/etc across the entire statistical spectrum) & age.

I expect to have a working version of the model done within the week.

J.E. wrote:the paper uses Berri's Wins Produced to gauge a player's production

That was a mistake.

J.E. wrote:The challenge is only open to residents of the U.S.

That leaves me out. I'll be rooting for people of this forum, then

I'm in, currently in the process of trying to streamline about 40 days left of draft model work I need to do into about 8 days.

Maybe Maymin will consider alternate WinScore (which, with the further adjustment to defensive rebounding, has performed well on some earlier tests I have seen) or another metric in the future.

I don't have time at this moment to compare the weights for components of alternate WinScore to other advanced SPMs out there. Has someone else does this or is interested in doing such a quick comparison? If so, thanks.


I think the Kings only have a first rounder and are certain to pick no worse than 7th? I don't have a lot of time to do this challenge right now. My initial thoughts were that if they got #1 or #2 pick they should rely on their scouting department's months / years of work on Wiggins and Embid and not really make it based on stats so much. If they didn't get #1 or #2, my gut says try to trade the pick for a vet or for pick #10-20 or some combination. I am not that impressed at first blush with who is available from 3-15 and think one might get more value, more fit with something else. Of the guys currently rated 15-25 at draft express there are some decent wings. I'd compare those guys especially on 3pt FG% and frequency, defense (steals, blocks and what shot defense info one could pull, especially from games against quality opponents). I'd look at PFs too.

Ultimately the pick strategy should flow from the overall strategy of how they want to play, what goals they have for team offensive and defensive efficiency. If one looked at the current data and maybe adjusted it somewhat for player development, you could then sub in a player into that rotation with targeted performance levels for your pick or acquisition as a starter or bench player and see how much that would contribute to achieving your goals. In setting those goals I'd look at the factor level data for the best 5 teams in the west overall and by position. I'd think the intermediate goal should be to design a team pretty close to the average strength levels of those teams / players. How many of their current starters are up to those averages? I'd guess it to be 1-2 but haven't done the work yet. But if I do have a little time I might do it this way.


Anyone else not planning to submit privately want to talk about the approach that makes sense to them publicly for the heck of it?

Anyone have college data they can share with me or post a link where I can find some data?

Thanks.

Have you checked kenpom.com, draftexpress.com, basketballreference.com? Those are good spots for data. I am not aware of better general databases for the college level, except I assume Synergy has college stats too?

J.E. wrote:This might be interesting to some.

The Kings will seek analysis on players from the public and eventually select two people to be in their draft room when they make their pick on June 26. Essentially the Kings are looking for amateurs to offer their analytics ideas to the team for free and in return they will be granted access

The team has launched a website where would be analysts will be asked to give rankings of draft prospects and their statistical methodology. Contestants are to use information from publicly available statistical databases to create their systems.

Link to the challenge

And now there's this.

NBA.COM wrote:The Kings put the No. 8 pick in the draft on the trade block immediately after the lottery Tuesday night

I don't think that's going to sit well with all the people who put effort into the contest. Although the article also says

It was immediately unclear if the Kings have a specific target. But Dennis Schroder, the German point guard firmly behind Jeff Teague on the Hawks depth chart, could come up

Given how extremely hilarious (and awful) that trade would be for the Kings one should probably be questioning the contents of the entire article