Page 2 of 2
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 2:43 pm
by DSMok1
Wonderful data, J.E.! Will be very useful.
Are you using BBRef ID's throughout? Do you have some sort of player database linking BBRef IDs to names and Bio data? Just asking.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 7:55 pm
by colts18
Can you explain the prior-informed RAPM and how it works? In the non-informed data, Duncan leads from 01-07 every single year, yet KG's 2003 and 2004 prior-informed is higher than Duncan's. If you are using prior data, wouldn't Duncan be higher since he is higher in every single season?
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 7:59 pm
by sideshowbob
Awesome, as usual.
Does this mean that BBV's 2012 data is complete, and that the prior-informed 2012 data should be up soon?
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 8:23 pm
by J.E.
DSMok1 wrote:Are you using BBRef ID's throughout?
I plan to do that, yes
Do you have some sort of player database linking BBRef IDs to names and Bio data?
I'll create files which have playerID and various BoxScore stats in csv format, and I have one file that has "bbr playerID;full name", but nothing with Bio data. What would you want to use that for?
Can you explain the prior-informed RAPM and how it works? In the non-informed data, Duncan leads from 01-07 every single year, yet KG's 2003 and 2004 prior-informed is higher than Duncan's. If you are using prior data, wouldn't Duncan be higher since he is higher in every single season?
There's a discrepancy here because the data I used for computing informed RAPM was incomplete. For most of those early years it was missing >15%
Does this mean that BBV's 2012 data is complete, and that the prior-informed 2012 data should be up soon?
?
Everything in this thread was grabbed from BBR. I don't think bbv has updated their 2012 matchupfile(s). 2012 informed RAPM is already up, it's just missing the playoffs. Hopefully I get to test all RAPM versions (vanilla, RAPM informed RAPM, BoxScore+PBP informed RAPM) before the season and then I'll upload whatever did best
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 8:33 pm
by sideshowbob
J.E. wrote:?
Everything in this thread was grabbed from BBR. I don't think bbv has updated their 2012 matchupfile(s). 2012 informed RAPM is already up, it's just missing the playoffs. Hopefully I get to test all RAPM versions (vanilla, RAPM informed RAPM, BoxScore+PBP informed RAPM) before the season and then I'll upload whatever did best
My mistake, I meant with the playoffs, and somehow it flew over my head that this was all done with BBR data. Just checked BBV, and yeah the data still hasn't been updated to include the playoffs.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Fri Sep 28, 2012 8:46 pm
by DSMok1
J.E. wrote:DSMok1 wrote:Are you using BBRef ID's throughout?
I plan to do that, yes
Do you have some sort of player database linking BBRef IDs to names and Bio data?
I'll create files which have playerID and various BoxScore stats in csv format, and I have one file that has "bbr playerID;full name", but nothing with Bio data. What would you want to use that for?
I thought it might be nice to link in with these tables for exact age data:
http://www.basketball-reference.com/players/a/
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Sat Sep 29, 2012 2:19 am
by colts18
For the missed quarters, couldn't you use ESPN or B-R's play by play to figure out who was playing in the missing time?
Is 2001 the last season you have pbp for? Is there anyone out there that has 2000 pbp data? I would bet that Shaq in 2000 might be the only guy with a 10+ RAPM.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Sat Sep 29, 2012 7:46 am
by J.E.
J.E. wrote:Everything in this thread was grabbed from BBR
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Wed Oct 03, 2012 10:10 pm
by J.E.
http://stats-for-the-nba.appspot.com/PB ... br_ids.rar
Contains (almost) all matchupdata from 2000/2001 to 2012, split into regular season and playoffs.
Fixed a bug where players that were substituted in during free throws showed up one possession too early in the matchupfiles.
From 2007 onwards it's pretty much bbv's dataset, with bbr player page urls listed instead of bbv's player id's. The games which bbv did not have (but bbr does) are at the bottom of each matchupfile.
For the year 2007 I'm torn between using bbv's matchupdata or mine. Mine probably has more errors, but bbv back then only listed total possessions for both teams, instead of splitting into home/away
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Thu Oct 04, 2012 12:09 am
by Crow
Is non-prior informed RAPM coming eventually for 2007-11?
Is there something that could done systematically with the prior-informed, non-prior informed and multiyear RAPM data over the stretch of time they are available to identify the most "out of step values", which might be signs of larger than average errors for those datapoints? I think that might be interesting and helpful.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Thu Oct 04, 2012 8:17 am
by J.E.
Is non-prior informed RAPM coming eventually for 2007-11?
Probably not
Is there something that could done systematically with the prior-informed, non-prior informed and multiyear RAPM data over the stretch of time they are available to identify the most "out of step values", which might be signs of larger than average errors for those datapoints?
What exactly do you mean by "out of step"?
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Thu Oct 04, 2012 1:04 pm
by DSMok1
Where do you explain the columns in the matchup files? I see no key.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Thu Oct 04, 2012 1:40 pm
by J.E.
It's the same format as bbv with the difference that bbv's files contain more information (that is not needed for RAPM or whatever).
If you split each line by TAB, gameid is at 0, home players are 5-9, away players are 10-14, home points #29, away points #30, home possessions #33, away possessions #34
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Thu Oct 04, 2012 5:10 pm
by Crow
Out of step could mean if one of the prior-informed, non-prior informed and multiyear RAPM values for a player showed X standard deviation more variance from the other values for that player than the average variance among them for those datasets.
Re: 2000-2006 PBP, matchupfile & RAPM
Posted: Tue Dec 04, 2012 6:47 pm
by DSMok1
J.E.: where did the matchup files go on your site?