APBRmetrics

Most people reading this forum will have already seen this, but I feel it's worth linking to again to make sure.

Tangotiger, one of the leading Sabermetrics experts, has been leading a discussion on his site on how to estimate replacement level for the NBA. He was one of the originators of the whole WAR replacement level concept for baseball, and is an expert on methodologies for estimating replacement level.

The thread is at: http://tangotiger.com/index.php/site/co ... nba-player

Good post on the topic:

I don’t think that RAPM is good for this type of thing because it regresses low minute (replacement level) players a lot. Maybe someone with RAPM expertise should run an RAPM with the 40 10 day contract players as just 1 variable rather than 40 seperate variables.

That would definitely be an interesting RAPM experiment. Put all the players who signed midseason (10 day, veterans, buyout players, etc.) into 1 variable and see what kind of RAPM they would have.

colts18 wrote:That would definitely be an interesting RAPM experiment. Put all the players who signed midseason (10 day, veterans, buyout players, etc.) into 1 variable and see what kind of RAPM they would have.

if someone wants to partner on this, then id be willing to do the rapm part. i just don't want to invest any summertime manhours of my team into researching the list of players (10 day, veterans, buyout players, etc.) over the past x years to then run the model with. but if someone is wiling to think through, and then put together, that list of guys (going back as far as 2006), then we can lump them together into a 'replacement level group' and run it. that part of the job is pretty trivial.

ftr weve done different but similar things, and my guess is still in the -2 to -2.5 or so range.

caveat: ive not read one word of the tango thread (yet), and thus far only know what ive seen about it on twitter (mostly via DSMok1).

If I don't read it here, or find a link from this forum, I probably didn't read it.

talkingpractice wrote:
if someone wants to partner on this, then id be willing to do the rapm part. i just don't want to invest any summertime manhours of my team into researching the list of players (10 day, veterans, buyout players, etc.) over the past x years to then run the model with. but if someone is wiling to think through, and then put together, that list of guys (going back as far as 2006), then we can lump them together into a 'replacement level group' and run it. that part of the job is pretty trivial.

ftr weve done different but similar things, and my guess is still in the -2 to -2.5 or so range.

caveat: ive not read one word of the tango thread (yet), and thus far only know what ive seen about it on twitter (mostly via DSMok1).

I don't have the data for past seasons, but here is a list of the players who were 10 day contract guys this past season. You can try to run a 2014 RAPM on them though I'm not sure if its a big enough sample (7406 total MP), but its a good start.

Drew Gooden
Chris Johnson
Cartier Martin
Jarvis Varnado
Shawne Williams
Dewayne Dedmon
Leandro Barbosa
Hilton Armstrong
Earl Clark
Chris Wright
Jorge Gutierrez
Casper Ware
Jason Collins
Jared Cunningham
Justin Hamilton
Othyus Jeffers
Orlando Johnson
Tony Mitchell
Maalik Wayns
Chris Babb
Seth Curry
Shane Edwards
Manny Harris
Royal Ivey
Damion James
DeAndre Liggins
Darius Morris
Arinze Onuaku
Dexter Pittman
D.J. Stephens
Adonis Thomas
D.J. White
Royce White
Reggie Williams
Vander Blue
Mike James
James Nunnally
Mustafa Shakur
Sasha Vujacic
Darius Johnson-Odom
Shannon Brown

Here is a list of minimum-salary players signed either before or during the season (noted as such) from 2008-09 through 2012-13: https://docs.google.com/spreadsheets/d/ ... xU/pubhtml

Excellent, Kevin! Thanks!

Well, that is perfect. I can get this done for us next week. I'll get one of our junior guys to handle any naming discrepancies between this list and our pbp, and then we can run it on the entire 2009-2014 sample (combining the lists from kevin and colts). I'm pretty sure this sample will be large enough to get a useful result.

Cross posting from the Tangotiger thread:

Thanks to Kevin Pelton for the list of minimum salary players signed preseason and in season for 2009 through 2013 (5 seasons). http://www.apbr.org/metrics/viewtopic.p ... 248#p20248

Here are their ASPM ratings, all weighted by minutes played: Once again, it appears replacement level is somewhere around -2.

what is meant by the Preseason row?

one of my guys already confirmed that there are way less naming discrepancies between these lists and our pbp than i guessed there would be originally, so ill have the rapm version of this back much quicker than i said earlier.

talkingpractice wrote:what is meant by the Preseason row?

one of my guys already confirmed that there are way less naming discrepancies between these lists and our pbp than i guessed there would be originally, so ill have the rapm version of this back much quicker than i said earlier.

Signed before the season vs. Signed during the season.

aaaaaaaaaaaahhhhhhhh, of course. im a bozo.

talkingpractice wrote:aaaaaaaaaaaahhhhhhhh, of course. im a bozo.

No you're not...

Someone on the Tangotiger thread noted that replacement level players are different before the season than during the season. Before the season, the available players will be better. During the season, most of the replacement level players will be in Europe playing, and the available players will not likely be as good.

One of the reasons for the wide disparity in the table I posted above.

That's what I was thinking. There are some decent role players before the season starts, who just missed the cut somewhere, due to roster makeup and other happenstance.

On colts18's list for 2013-14, we have Cartier Martin, who had started 6 games in 3 years for the Wiz. He had .065 WS/48 in that interval, which is much better than I normally think of as guy-off-the-street caliber. He played 12% of the ~7500 total minutes by these players last year.

I get 6.15 eWins for the lot of them -- a rate that is .40 of NBA avg.
Normally I consider 'replacement value' to achieve zero wins. These guys represent about 40% of the bunch -- and 105% of the eWins total. The rest had few minutes or were more truly off-the-street performers.

i'll have the rapm version of this for us by Monday at the latest.