1-Year NPI APM

[Nathan]

Hi All,

I've been doing some work making a statistical plus/minus (and i'm hoping to make a draft rater along the same lines) using JE's previous year informed RAPM. However, I have some worries about overfitting because data points from consecutive years are far from independent (and, relatedly, I'm having a hard time coming up with estimated uncertainties in my ratings). So, I'm looking for non-prior informed APM, preferably in 1-year samples over the course of at least 10 years, and preferably including the 2014-15 season, but I realize I may have to settle on at least one of those counts. What are my options?

Thanks! And, if anyone's curious, here are the players who rate +3 or better for the 2013-14 season (out of sample) in my current model, (pure box score, no plus/minus information)

[Crow]

multi-yr NPI option here http://shutupandjam.net/

some 1yr stuff here https://sites.google.com/site/rapmstats/ including 2012.

Might check for back threads here too. Or hope JE is willing to assist directly.

[Nathan]

Thanks for the advice! I guess I should also ask (although this seems less likely), is there publicly available "stints" data, from which I might calculate my own APM?

[Crow]

Pay option- Play by play near bottom https://downloads.nbastuffer.com/?ap_id=nbastuffer but you might have to form the lineup stints. Ask to see a file sample or tell him your concerns / must haves.

Read APM related threads and maybe pm some of the folks who appear to have such sets. Evan Zamir at nbawowy.com might be one possibility. Pay or gratis, up to them.

Basketballvalue.com still has downloads available http://basketballvalue.com/downloads.php for up thru 2011-12. If you are interested in APM as an intermediate resource maybe it would do.

[Nathan]

Awesome, thanks so much for the help!

[J.E.]

viewtopic.php?p=24708#p24708

[Nathan]

Ooooh, this is exactly what I was looking for. Thanks, JE.

One comment: I noticed this in the 2001 data and in the 2002 data



What player is this supposed to be? I'll keep an eye out for any similar occurrences.

[Statman]

Ooooh, this is exactly what I was looking for. Thanks, JE.

One comment: I noticed this in the 2001 data and in the 2002 data



What player is this supposed to be? I'll keep an eye out for any similar occurrences.

Funny, because according to B-R no past NBA player has that birth date.

[J.E.]

Ooooh, this is exactly what I was looking for. Thanks, JE.

One comment: I noticed this in the 2001 data and in the 2002 data



What player is this supposed to be? I'll keep an eye out for any similar occurrences.

It's Sam Mitchell. Not entirely sure what happened there

[Nathan]

Ok, that makes sense, I think. Tyrone Corbin also had his name replaced by a date in the 2001 data. I'm pretty sure those are the only two such cases.

Thanks again!