Author Message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Dec 29, 2010 9:33 pm Post subject: ezPM Reply with quote
I've got PBP data in my linear weights model now. Thought some of you would be interested. I'd go into depth here, but I think most of it is old hat to you guys. If you have any questions/comments/suggestions, feel free to leave them here or at my blog.
http://thecity2.com/2010/12/28/ezpm-1-0 ... play-data/
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC
PostPosted: Thu Dec 30, 2010 12:27 pm Post subject: Reply with quote
Many fans will have a problem with this language:
Quote:
Say Curry was +5, Ellis was +6, Lee was +2, Biedrins was +1 and Dorell Wright was -4. In this example, the team +/- is +10, but Dorell Wright was actually a negative contributer.
Unless this covered relatively few minutes, can you really say Wright made negative overall contributions?
Put more broadly, does it make real-world sense to say the entirety of NBA rosters make zero net contributions?
In the NBA Finals, with the 2 best teams in the world going at it, are there as many negative as positive acts?
If the Warriors should bench all their 'negative contributors', would they win more games? Would they have 5 players?
Do you really win 41 games with zero net contribution toward winning those many games?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Thu Dec 30, 2010 12:37 pm Post subject: Reply with quote
Mike G, that's the definition of plus/minus.
Your comments are why replacement level is required to value players.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1
Back to top
View user's profile Send private message Send e-mail Visit poster's website
bbstats
Joined: 25 Apr 2010
Posts: 46
PostPosted: Thu Dec 30, 2010 12:50 pm Post subject: Reply with quote
For clarification: to be .500 is to win by 0, on average.
_________________
http://thebasketballdistribution.blogspot.com
http://twitter.com/bbstats
Back to top
View user's profile Send private message Visit poster's website
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Thu Dec 30, 2010 1:11 pm Post subject: Reply with quote
Mike G wrote:
If the Warriors should bench all their 'negative contributors', would they win more games? Would they have 5 players?
Uh, I really think if they benched Vlad Rad, they would have won at least 2 more games. Laughing
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Thu Dec 30, 2010 1:21 pm Post subject: Reply with quote
EvanZ wrote:
Mike G wrote:
If the Warriors should bench all their 'negative contributors', would they win more games? Would they have 5 players?
Uh, I really think if they benched Vlad Rad, they would have won at least 2 more games. Laughing
ASPM has Vlad Rad as close to league-average. What's the big difference between your numbers and mine on him? I have him as a -1.2 offensive player (his numbers aren't that bad) and a +1.2 defender (the best on Golden State). The defense is based on his high steal and block numbers. What do you see differently?
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1
Back to top
View user's profile Send private message Send e-mail Visit poster's website
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Thu Dec 30, 2010 1:27 pm Post subject: Reply with quote
My model has him at -4.81 (-1.5/-3.3). My eyes tend to agree. He's a nightmare on offense and defense most of the time. Can't stay in front of anyone. He hits a three every once in a while, and is a tall white European guy (who had a few good playoff games with LAL). Somehow, Keith Smart trusts him.
BTW, not just my opinion. If you ever go over to GSOM, he is universally despised.
EDIT: I'm referring to 2009-10 stats. I haven't crunched this season yet.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
BobboFitos
Joined: 21 Feb 2009
Posts: 200
Location: Cambridge, MA
PostPosted: Thu Dec 30, 2010 3:09 pm Post subject: Reply with quote
DSMok1 wrote:
EvanZ wrote:
Mike G wrote:
If the Warriors should bench all their 'negative contributors', would they win more games? Would they have 5 players?
Uh, I really think if they benched Vlad Rad, they would have won at least 2 more games. Laughing
ASPM has Vlad Rad as close to league-average. What's the big difference between your numbers and mine on him? I have him as a -1.2 offensive player (his numbers aren't that bad) and a +1.2 defender (the best on Golden State). The defense is based on his high steal and block numbers. What do you see differently?
Counterpart data suggests Vlad Rad is a +3.3 per 100 defender. (Meaning, he's 3.3 points below average) From the eye test as well/stereotype about older tall Euros, I'm prone to believe that. Take it for whatever it's worth though, counterpart isn't reliable etc etc.
_________________
http://pointsperpossession.com/
@PPPBasketball
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
PostPosted: Thu Dec 30, 2010 3:12 pm Post subject: Reply with quote
He's unusual in that he has high steal rates and high block rates.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1
Back to top
View user's profile Send private message Send e-mail Visit poster's website
BobboFitos
Joined: 21 Feb 2009
Posts: 200
Location: Cambridge, MA
PostPosted: Thu Dec 30, 2010 3:14 pm Post subject: Reply with quote
DSMok1 wrote:
He's unusual in that he has high steal rates and high block rates.
He's also way outperforming his career norms on blocks and steals this year. Per 36, he's 2.0 spg/1.2 bpg. Career wise, 1.2 and .5, respectively.
_________________
http://pointsperpossession.com/
@PPPBasketball
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Thu Dec 30, 2010 3:16 pm Post subject: Reply with quote
DSMok1 wrote:
He's unusual in that he has high steal rates and high block rates.
Yeah, that's true. Interestingly, though, ezPM places quite a bit of weight on blocks (0.7). He would be much worse off without it I guess. Like I said before, I need to do this season. Maybe my eyes are deceiving me.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Fri Dec 31, 2010 8:05 pm Post subject: Reply with quote
I've done the 2010-11 numbers through Wednesday:
Code:
Rank ORank DRank LAST FIRST TEAM POS EZPM100 OFF100 DEF100 POSS
1 1 3 Paul Chris NOH 1 11.02 8.93 2.10 1833
2 10 7 Wade Dwyane MIA 2 6.73 5.48 1.24 2125
3 18 5 James LeBron MIA 3 6.67 5.20 1.47 2304
4 2 114 Nash Steve PHX 1 6.45 8.08 -1.63 1788
5 3 84 Williams Deron UTA 1 6.13 7.22 -1.09 2150
6 6 26 Ginobili Manu SAS 2 5.97 5.92 0.05 1871
7 97 1 Garnett Kevin BOS 4 5.29 2.16 3.13 1698
8 16 25 Rose Derrick CHI 1 5.28 5.24 0.05 1938
9 9 58 Parker Tony SAS 1 4.86 5.51 -0.65 1859
10 89 2 Howard Dwight ORL 5 4.86 2.36 2.50 1834
11 42 13 Rondo Rajon BOS 1 4.70 3.89 0.80 1443
12 29 31 Carter Vince ORL 2 4.38 4.47 -0.09 1114
13 11 86 Horford Al ATL 5 4.37 5.47 -1.11 1918
14 12 91 Gasol Pau LAL 5 4.23 5.44 -1.21 2250
15 15 82 Nowitzki Dirk DAL 4 4.18 5.26 -1.09 1797
16 33 32 Miller Andre POR 1 4.16 4.31 -0.14 1695
17 24 56 Westbrook Russell OKC 1 4.07 4.69 -0.62 2076
18 34 36 Pierce Paul BOS 3 4.02 4.26 -0.23 1946
19 44 24 Young Thaddeus PHI 3 3.98 3.84 0.14 1616
20 5 140 Love Kevin MIN 4 3.92 5.97 -2.05 2139
21 20 76 Bryant Kobe LAL 2 3.91 4.93 -1.02 1850
22 7 133 Nelson Jameer ORL 1 3.86 5.84 -1.99 1407
23 22 80 Felton Raymond NYK 1 3.82 4.88 -1.06 2266
24 35 50 Chandler Tyson DAL 5 3.74 4.23 -0.48 1287
25 4 165 Martin Kevin HOU 2 3.61 6.17 -2.57 1823
26 83 9 Kidd Jason DAL 1 3.56 2.46 1.11 1645
27 88 8 Duncan Tim SAS 5 3.56 2.36 1.20 1602
28 48 38 Anthony Carmelo DEN 3 3.36 3.62 -0.26 1556
29 19 126 Harris Devin NJN 1 3.34 5.15 -1.82 1548
30 13 147 Hilario Nene DEN 5 3.27 5.44 -2.17 1561
31 51 35 Lowry Kyle HOU 1 3.17 3.39 -0.23 1657
32 41 61 Noah Joakim CHI 5 3.17 3.90 -0.74 1545
33 57 29 Brown Shannon LAL 2 3.15 3.18 -0.03 1277
34 54 33 Hill George SAS 1 3.13 3.28 -0.15 1540
35 131 4 Brewer Ronnie CHI 2 3.10 1.34 1.76 1278
36 72 23 Gay Rudy MEM 3 3.07 2.77 0.30 2249
37 30 98 Odom Lamar LAL 4 3.04 4.44 -1.39 2054
38 14 157 Gordon Eric LAC 2 2.98 5.39 -2.41 1987
39 50 57 Durant Kevin OKC 3 2.97 3.60 -0.63 2083
40 26 112 Calderon Jose TOR 1 2.96 4.56 -1.60 1330
41 81 18 Thomas Tyrus CHA 4 2.94 2.47 0.47 1010
42 31 101 Randolph Zach MEM 4 2.87 4.35 -1.48 1817
43 40 83 Brand Elton PHI 5 2.85 3.93 -1.09 1717
44 28 119 Griffin Blake LAC 4 2.82 4.55 -1.72 2037
45 59 47 Allen Ray BOS 2 2.77 3.16 -0.39 1969
46 65 40 Barnes Matt LAL 3 2.74 3.04 -0.30 1438
47 99 16 Iguodala Andre PHI 2 2.72 2.06 0.65 1533
48 68 46 Bosh Chris MIA 4 2.59 2.96 -0.37 2181
49 27 137 Lawson Ty DEN 1 2.55 4.55 -2.00 1740
50 108 14 McGrady Tracy DET 3 2.52 1.84 0.69 1058
Vince Carter! Shocked
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
deepak
Joined: 26 Apr 2006
Posts: 665
PostPosted: Fri Dec 31, 2010 11:08 pm Post subject: Reply with quote
EvanZ wrote:
I've done the 2010-11 numbers through Wednesday:
Vince Carter! Shocked
Thanks. Its interesting to me that everyone in the top 50 is a positive offensive contributor, while most are negative defensive contributors. Will the Def100 for the entire league sum up to 0?
Here it is, with easier to read formatting:
Code:
Rank ORank DRank Name TEAM POS EZPM100 OFF100 DEF100 POSS
1 1 3 Chris Paul NOH 1 11.02 8.93 2.1 1833
2 10 7 Dwyane Wade MIA 2 6.73 5.48 1.24 2125
3 18 5 LeBron James MIA 3 6.67 5.2 1.47 2304
4 2 114 Steve Nash PHX 1 6.45 8.08 -1.63 1788
5 3 84 Deron Williams UTA 1 6.13 7.22 -1.09 2150
6 6 26 Manu Ginobili SAS 2 5.97 5.92 0.05 1871
7 97 1 Kevin Garnett BOS 4 5.29 2.16 3.13 1698
8 16 25 Derrick Rose CHI 1 5.28 5.24 0.05 1938
9 9 58 Tony Parker SAS 1 4.86 5.51 -0.65 1859
10 89 2 Dwight Howard ORL 5 4.86 2.36 2.5 1834
11 42 13 Rajon Rondo BOS 1 4.7 3.89 0.8 1443
12 29 31 Vince Carter ORL 2 4.38 4.47 -0.09 1114
13 11 86 Al Horford ATL 5 4.37 5.47 -1.11 1918
14 12 91 Pau Gasol LAL 5 4.23 5.44 -1.21 2250
15 15 82 Dirk Nowitzki DAL 4 4.18 5.26 -1.09 1797
16 33 32 Andre Miller POR 1 4.16 4.31 -0.14 1695
17 24 56 Russell Westbrook OKC 1 4.07 4.69 -0.62 2076
18 34 36 Paul Pierce BOS 3 4.02 4.26 -0.23 1946
19 44 24 Thaddeus Young PHI 3 3.98 3.84 0.14 1616
20 5 140 Kevin Love MIN 4 3.92 5.97 -2.05 2139
21 20 76 Kobe Bryant LAL 2 3.91 4.93 -1.02 1850
22 7 133 Jameer Nelson ORL 1 3.86 5.84 -1.99 1407
23 22 80 Raymond Felton NYK 1 3.82 4.88 -1.06 2266
24 35 50 Tyson Chandler DAL 5 3.74 4.23 -0.48 1287
25 4 165 Kevin Martin HOU 2 3.61 6.17 -2.57 1823
26 83 9 Jason Kidd DAL 1 3.56 2.46 1.11 1645
27 88 8 Tim Duncan SAS 5 3.56 2.36 1.2 1602
28 48 38 Carmelo Anthony DEN 3 3.36 3.62 -0.26 1556
29 19 126 Devin Harris NJN 1 3.34 5.15 -1.82 1548
30 13 147 Nene Hilario DEN 5 3.27 5.44 -2.17 1561
31 51 35 Kyle Lowry HOU 1 3.17 3.39 -0.23 1657
32 41 61 Joakim Noah CHI 5 3.17 3.9 -0.74 1545
33 57 29 Shannon Brown LAL 2 3.15 3.18 -0.03 1277
34 54 33 George Hill SAS 1 3.13 3.28 -0.15 1540
35 131 4 Ronnie Brewer CHI 2 3.1 1.34 1.76 1278
36 72 23 Rudy Gay MEM 3 3.07 2.77 0.3 2249
37 30 98 Lamar Odom LAL 4 3.04 4.44 -1.39 2054
38 14 157 Eric Gordon LAC 2 2.98 5.39 -2.41 1987
39 50 57 Kevin Durant OKC 3 2.97 3.6 -0.63 2083
40 26 112 Jose Calderon TOR 1 2.96 4.56 -1.6 1330
41 81 18 Tyrus Thomas CHA 4 2.94 2.47 0.47 1010
42 31 101 Zach Randolph MEM 4 2.87 4.35 -1.48 1817
43 40 83 Elton Brand PHI 5 2.85 3.93 -1.09 1717
44 28 119 Blake Griffin LAC 4 2.82 4.55 -1.72 2037
45 59 47 Ray Allen BOS 2 2.77 3.16 -0.39 1969
46 65 40 Matt Barnes LAL 3 2.74 3.04 -0.3 1438
47 99 16 Andre Iguodala PHI 2 2.72 2.06 0.65 1533
48 68 46 Chris Bosh MIA 4 2.59 2.96 -0.37 2181
49 27 137 Ty Lawson DEN 1 2.55 4.55 -2 1740
50 108 14 Tracy McGrady DET 3 2.52 1.84 0.69 1058
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Fri Dec 31, 2010 11:27 pm Post subject: Reply with quote
deepak wrote:
Thanks. Its interesting to me that everyone in the top 50 is a positive offensive contributor, while most are negative defensive contributors. Will the Def100 for the entire league sum up to 0?
Defense + offense should theoretically sum to zero in my model. Here are the actual totals for the entire league:
#Possessions (all players summed): 408,523
Offensive ezPM points: +7,033
Defensive ezPM points: -6,691
Total (Off+Def)/#Possessions: uh, really close to zero obviously Wink
BTW, how do you get the formatting so nice? Not by hand, I assume.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
deepak
Joined: 26 Apr 2006
Posts: 665
PostPosted: Fri Dec 31, 2010 11:41 pm Post subject: Reply with quote
EvanZ wrote:
BTW, how do you get the formatting so nice? Not by hand, I assume.
I copied it into Excel (text to columns), merged the first and last name columns. Then copied it into my text editor (vim), replaced tabs with spaces, and adjusted the spacing between the columns (easy to do in vim).
Author Message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Fri Dec 31, 2010 11:54 pm Post subject: Reply with quote
deepak wrote:
EvanZ wrote:
BTW, how do you get the formatting so nice? Not by hand, I assume.
I copied it into Excel (text to columns), merged the first and last name columns. Then copied it into my text editor (vim), replaced tabs with spaces, and adjusted the spacing between the columns (easy to do in vim).
ah, almost by hand. Laughing
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 12:07 am Post subject: Reply with quote
Here's the full spreadsheet with all players and all the raw stats (.csv), if anyone wants it.
https://spreadsheets.google.com/pub?key ... output=csv
Code:
Rank ORank DRank Name TEAM POS EZPM100 OFF100 DEF100 POSS
1 1 3 Chris Paul NOH 1 11.02 8.93 2.1 1833
2 10 7 Dwyane Wade MIA 2 6.73 5.48 1.24 2125
3 18 5 LeBron James MIA 3 6.67 5.2 1.47 2304
4 2 114 Steve Nash PHX 1 6.45 8.08 -1.63 1788
5 3 84 Deron Williams UTA 1 6.13 7.22 -1.09 2150
6 6 26 Manu Ginobili SAS 2 5.97 5.92 0.05 1871
7 97 1 Kevin Garnett BOS 4 5.29 2.16 3.13 1698
8 16 25 Derrick Rose CHI 1 5.28 5.24 0.05 1938
9 9 58 Tony Parker SAS 1 4.86 5.51 -0.65 1859
10 89 2 Dwight Howard ORL 5 4.86 2.36 2.5 1834
11 42 13 Rajon Rondo BOS 1 4.7 3.89 0.8 1443
12 29 31 Vince Carter ORL 2 4.38 4.47 -0.09 1114
13 11 86 Al Horford ATL 5 4.37 5.47 -1.11 1918
14 12 91 Pau Gasol LAL 5 4.23 5.44 -1.21 2250
15 15 82 Dirk Nowitzki DAL 4 4.18 5.26 -1.09 1797
16 33 32 Andre Miller POR 1 4.16 4.31 -0.14 1695
17 24 56 Russell Westbrook OKC 1 4.07 4.69 -0.62 2076
18 34 36 Paul Pierce BOS 3 4.02 4.26 -0.23 1946
19 44 24 Thaddeus Young PHI 3 3.98 3.84 0.14 1616
20 5 140 Kevin Love MIN 4 3.92 5.97 -2.05 2139
21 20 76 Kobe Bryant LAL 2 3.91 4.93 -1.02 1850
22 7 133 Jameer Nelson ORL 1 3.86 5.84 -1.99 1407
23 22 80 Raymond Felton NYK 1 3.82 4.88 -1.06 2266
24 35 50 Tyson Chandler DAL 5 3.74 4.23 -0.48 1287
25 4 165 Kevin Martin HOU 2 3.61 6.17 -2.57 1823
26 83 9 Jason Kidd DAL 1 3.56 2.46 1.11 1645
27 88 8 Tim Duncan SAS 5 3.56 2.36 1.2 1602
28 48 38 Carmelo Anthony DEN 3 3.36 3.62 -0.26 1556
29 19 126 Devin Harris NJN 1 3.34 5.15 -1.82 1548
30 13 147 Nene Hilario DEN 5 3.27 5.44 -2.17 1561
31 51 35 Kyle Lowry HOU 1 3.17 3.39 -0.23 1657
32 41 61 Joakim Noah CHI 5 3.17 3.9 -0.74 1545
33 57 29 Shannon Brown LAL 2 3.15 3.18 -0.03 1277
34 54 33 George Hill SAS 1 3.13 3.28 -0.15 1540
35 131 4 Ronnie Brewer CHI 2 3.1 1.34 1.76 1278
36 72 23 Rudy Gay MEM 3 3.07 2.77 0.3 2249
37 30 98 Lamar Odom LAL 4 3.04 4.44 -1.39 2054
38 14 157 Eric Gordon LAC 2 2.98 5.39 -2.41 1987
39 50 57 Kevin Durant OKC 3 2.97 3.6 -0.63 2083
40 26 112 Jose Calderon TOR 1 2.96 4.56 -1.6 1330
41 81 18 Tyrus Thomas CHA 4 2.94 2.47 0.47 1010
42 31 101 Zach Randolph MEM 4 2.87 4.35 -1.48 1817
43 40 83 Elton Brand PHI 5 2.85 3.93 -1.09 1717
44 28 119 Blake Griffin LAC 4 2.82 4.55 -1.72 2037
45 59 47 Ray Allen BOS 2 2.77 3.16 -0.39 1969
46 65 40 Matt Barnes LAL 3 2.74 3.04 -0.3 1438
47 99 16 Andre Iguodala PHI 2 2.72 2.06 0.65 1533
48 68 46 Chris Bosh MIA 4 2.59 2.96 -0.37 2181
49 27 137 Ty Lawson DEN 1 2.55 4.55 -2 1740
50 108 14 Tracy McGrady DET 3 2.52 1.84 0.69 1058
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 584
Location: Where the wind comes sweeping down the plains
PostPosted: Sat Jan 01, 2011 10:12 am Post subject: Reply with quote
EvanZ, I created an Excel spreadsheet that will automatically generate formatted text for this site. It's here:
https://docs.google.com/leaf?id=0Bx1NfC ... ZWVi&hl=en
Just paste in the table you want, adjust the column widths as you desire, and export into a formatted plain text file.
I never got around to posting it on here (I only made it a few weeks ago) but I think it could save people a lot of difficulty!
_________________
GodismyJudgeOK.com/DStats
Back to top
View user's profile Send private message Visit poster's website
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 10:23 am Post subject: Reply with quote
DSMok1 wrote:
EvanZ, I created an Excel spreadsheet that will automatically generate formatted text for this site. It's here:
https://docs.google.com/leaf?id=0Bx1NfC ... ZWVi&hl=en
Just paste in the table you want, adjust the column widths as you desire, and export into a formatted plain text file.
I never got around to posting it on here (I only made it a few weeks ago) but I think it could save people a lot of difficulty!
Awesome. Thanks. I found a bug in my code that resulted in undercounting possessions. It should be fixed now. I'll post the new results soon.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 10:54 am Post subject: Reply with quote
Evan: Is there a reason you think total offense and total defense shouldn't each add to zero?
It looks to me like the player variance on offense is much larger than on defense. Do you think this is just a limitation of the data that we have, or that offensive variance really is much larger?
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 584
Location: Where the wind comes sweeping down the plains
PostPosted: Sat Jan 01, 2011 11:54 am Post subject: Reply with quote
Guy wrote:
Evan: Is there a reason you think total offense and total defense shouldn't each add to zero?
It looks to me like the player variance on offense is much larger than on defense. Do you think this is just a limitation of the data that we have, or that offensive variance really is much larger?
I think offensive variance is significantly larger (even at the team level, to some extent.) I also think our data is more limited on defense--so I think both factors are at work.
_________________
GodismyJudgeOK.com/DStats
Back to top
View user's profile Send private message Visit poster's website
Crow
Joined: 20 Jan 2009
Posts: 796
PostPosted: Sat Jan 01, 2011 12:52 pm Post subject: Reply with quote
Would you be interested in breaking the scoring defense (including free throws allowed) part of the overall formula out on its own as an additional separate column and showed the league-wide ranking for that since this is the most new element relative to other metrics? The points or fractions of points of this specific player level impact is estimated to have compared to a player's overall impact and total defensive impact would be helpful to see.
It might also be useful to see how much the correlation of your metric without scoring defense and Adjusted +/- changes when scoring defense is added. And perhaps also useful to see with and without comparisons to other metrics.
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 1:19 pm Post subject: Reply with quote
DS, I don't have Excel, so I wrote my own Ruby script. Anyway, here are the update 2010-11 data (through Friday's games) with possession count fixed.
Code:
SEASON OVERALL ORANK DRANK LAST FIRST TEAM POS #POSS EZPM100 O100 D100
2010-2011 1 1 3 Paul Chris NOH 1 2295 9.21 7.34 1.87
2010-2011 2 10 7 Wade Dwyane MIA 2 2418 5.91 4.82 1.09
2010-2011 3 13 5 James LeBron MIA 4 2603 5.91 4.61 1.30
2010-2011 4 2 106 Nash Steve PHX 1 2094 5.51 6.90 -1.39
2010-2011 5 3 88 Williams Deron UTA 1 2579 5.07 6.24 -1.17
2010-2011 6 7 24 Ginobili Manu SAS 2 2278 4.98 4.91 0.07
2010-2011 7 15 23 Rose Derrick CHI 1 2321 4.64 4.56 0.08
2010-2011 8 99 1 Garnett Kevin BOS 4 1999 4.49 1.83 2.66
2010-2011 9 86 2 Howard Dwight ORL 5 2102 4.19 2.03 2.16
2010-2011 10 9 61 Gasol Pau LAL 4 2624 4.19 4.85 -0.66
2010-2011 11 27 22 Young Thaddeus PHI 4 1559 4.13 3.98 0.15
2010-2011 12 42 13 Rondo Rajon BOS 1 1649 4.11 3.41 0.70
2010-2011 13 17 51 Parker Tony SAS 1 2264 4.05 4.49 -0.44
2010-2011 14 12 77 Horford Al ATL 5 2246 3.73 4.67 -0.94
2010-2011 15 31 30 Carter Vince ORL 2 1312 3.72 3.80 -0.08
2010-2011 16 32 37 Miller Andre POR 1 2075 3.53 3.78 -0.26
2010-2011 17 19 74 Nowitzki Dirk DAL 4 2177 3.45 4.34 -0.90
2010-2011 18 30 55 Westbrook Russell OKC 1 2471 3.42 3.94 -0.52
2010-2011 19 40 27 Brown Shannon LAL 2 1177 3.41 3.45 -0.04
2010-2011 20 41 29 Hill George SAS 1 1591 3.35 3.42 -0.08
2010-2011 21 23 73 Felton Raymond NYK 1 2599 3.34 4.22 -0.89
2010-2011 22 5 128 Love Kevin MIN 4 2517 3.33 5.07 -1.74
2010-2011 23 8 116 Hilario Nene DEN 5 1824 3.30 4.85 -1.55
2010-2011 24 39 33 Pierce Paul BOS 3 2298 3.27 3.48 -0.21
2010-2011 25 25 70 Bryant Kobe LAL 2 2232 3.24 4.08 -0.84
2010-2011 26 6 125 Nelson Jameer ORL 1 1655 3.23 4.92 -1.69
2010-2011 27 82 9 Duncan Tim SAS 5 1997 3.09 2.10 0.99
2010-2011 28 54 28 Boozer Carlos CHI 4 1057 3.09 3.14 -0.05
2010-2011 29 4 156 Martin Kevin HOU 2 2195 2.99 5.18 -2.19
2010-2011 30 90 8 Kidd Jason DAL 1 2099 2.99 1.97 1.02
2010-2011 31 53 35 Anthony Carmelo DEN 3 1789 2.92 3.15 -0.22
2010-2011 32 37 58 Brand Elton PHI 4 2066 2.91 3.49 -0.58
2010-2011 33 21 108 Calderon Jose TOR 1 1526 2.87 4.28 -1.41
2010-2011 34 52 42 Barnes Matt LAL 3 1386 2.84 3.15 -0.31
2010-2011 35 44 60 Noah Joakim CHI 5 1803 2.71 3.35 -0.63
2010-2011 36 75 20 Gay Rudy MEM 3 2562 2.69 2.43 0.26
2010-2011 37 56 52 Chandler Tyson DAL 5 1725 2.61 3.06 -0.45
2010-2011 38 14 146 Lawson Ty DEN 1 1724 2.57 4.59 -2.02
2010-2011 39 36 90 Odom Lamar LAL 4 2434 2.57 3.75 -1.18
2010-2011 40 57 56 Durant Kevin OKC 3 2449 2.52 3.06 -0.54
2010-2011 41 114 10 Camby Marcus POR 5 1643 2.49 1.58 0.92
2010-2011 42 33 103 Randolph Zach MEM 4 2095 2.49 3.77 -1.28
2010-2011 43 24 124 Harris Devin NJN 1 1904 2.48 4.17 -1.69
2010-2011 44 62 41 Lowry Kyle HOU 1 1969 2.46 2.77 -0.31
2010-2011 45 100 14 McGrady Tracy DET 3 1086 2.46 1.79 0.67
2010-2011 46 16 148 Gordon Eric LAC 2 2381 2.44 4.50 -2.06
2010-2011 47 61 45 Allen Ray BOS 2 2333 2.41 2.78 -0.38
2010-2011 48 26 121 Griffin Blake LAC 4 2447 2.38 4.03 -1.65
2010-2011 49 35 107 Curry Stephen GSW 1 1711 2.37 3.77 -1.39
2010-2011 50 22 142 Johnson Amir TOR 4 1601 2.30 4.27 -1.97
2010-2011 51 68 43 Bosh Chris MIA 5 2486 2.27 2.60 -0.33
If anyone want the script to make formatted text for the forum, here it is (Ruby is extremely easy to install anywhere):
Code:
input="/stats/ezPM Data/2010/ezpm_data_20101231.csv"
output="/stats/ezPM Data/2010/top50_20101231.txt"
$rows=IO.readlines(input)
$nice_out=File.new(output,"w")
$nice_out.puts "SEASON OVERALL ORANK DRANK LAST FIRST TEAM POS #POSS EZPM100 O100 D100"
1.upto(51) do
|i|
player=$rows.split(",")
col2=" " #OVERALL
col2[1..(player[1].size)]=player[1]
col3=" " #ORANK
col3[1..player[2].size]=player[2]
col4=" " #DRANK
col4[1..player[3].size]=player[3]
col5=" " #LAST
col5[1..player[4].size]=player[4]
col6=" " #FIRST
col6[1..player[5].size]=player[5]
col7=" " #TEAM
col7[1..player[6].size]=player[6]
col8=" " #POSITION
col8[1..player[7].size]=player[7]
col9=" " #POSSESSIONS
col9[1..player[8].size]=player[8]
col10=" " #EZPM100
col10[1..player[10].size]=player[10]
col11=" " #O100
col11[1..player[12].size]=player[12]
col12=" " #D100
col12[1..player[14].size]=player[14]
$nice_out.puts player[0]+" "+col2+col3+col4+col5+col6+col7+col8+col9+col10+col11+col12
end
$nice_out.close
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Last edited by EvanZ on Sat Jan 01, 2011 1:27 pm; edited 1 time in total
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 1:22 pm Post subject: Reply with quote
Crow wrote:
Would you be interested in breaking the scoring defense (including free throws allowed) part of the overall formula out on its own as an additional separate column and showed the league-wide ranking for that since this is the most new element relative to other metrics? The points or fractions of points of this specific player level impact is estimated to have compared to a player's overall impact and total defensive impact would be helpful to see.
It might also be useful to see how much the correlation of your metric without scoring defense and Adjusted +/- changes when scoring defense is added. And perhaps also useful to see with and without comparisons to other metrics.
Crow, here's the full spreadsheet, if you want to look at individual stats.
https://spreadsheets.google.com/pub?key ... output=csv
I'll definitely get to your questions as I have time.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 796
PostPosted: Sat Jan 01, 2011 1:57 pm Post subject: Reply with quote
Alright, thanks for the prompt. I probably should have realized that I can put what I want to see together but mentioning it brought that prompt and shared the idea.
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 2:02 pm Post subject: Reply with quote
Quote:
I think offensive variance is significantly larger (even at the team level, to some extent.) I also think our data is more limited on defense--so I think both factors are at work.
My sense is that team variance on defensive efficiency is very similar to offensive efficiency. Am I wrong about that? If I'm correct, then it would seem that the variance in player defense must also be as large. So I think that the appearance of a narrower range of defensive talent must mainly be a function of the data/models, rather than reality.
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 2:10 pm Post subject: Reply with quote
Here's my first crack at single game analysis for yesterday's GSW-CHA game. The ezPM for both teams sums to 1.5 pts. Pretty close to zero.
http://thecity2.com/2011/01/01/introduc ... 96-cha-95/
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 2:12 pm Post subject: Reply with quote
Guy wrote:
Quote:
I think offensive variance is significantly larger (even at the team level, to some extent.) I also think our data is more limited on defense--so I think both factors are at work.
My sense is that team variance on defensive efficiency is very similar to offensive efficiency. Am I wrong about that? If I'm correct, then it would seem that the variance in player defense must also be as large. So I think that the appearance of a narrower range of defensive talent must mainly be a function of the data/models, rather than reality.
Defensive variance is clearly smaller in my model, and that appears to be mostly a function of spreading the blame/credit for defense. Actually, I think that's a sanity check. As I change the defensive model to attempt to individualize credit, the variance will increase.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Ed Küpfer
Joined: 30 Dec 2004
Posts: 784
Location: Toronto
PostPosted: Sat Jan 01, 2011 5:05 pm Post subject: Reply with quote
Guy wrote:
My sense is that team variance on defensive efficiency is very similar to offensive efficiency. Am I wrong about that?
Team offense is slightly more variable than defense, using a number of different metrics and methods. I'm not aware of anybody that has found anything different. As far as I know, no one has found out why.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
EvanZ
Joined: 22 Nov 2010
Posts: 244
PostPosted: Sat Jan 01, 2011 5:29 pm Post subject: Reply with quote
Ed Küpfer wrote:
Guy wrote:
My sense is that team variance on defensive efficiency is very similar to offensive efficiency. Am I wrong about that?
Team offense is slightly more variable than defense, using a number of different metrics and methods. I'm not aware of anybody that has found anything different. As far as I know, no one has found out why.
Take two teams with the following ratings:
Code:
Team OFF DEF
A 105 100
B 100 95
If it's true that offense has more variance, Team B would win more, right?
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
EZPM (EvanZ, 2010)
EZPM (EvanZ, 2010)
Last edited by Crow on Thu May 12, 2011 6:43 am, edited 1 time in total.
Re: EZPM
Author Message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sat Jan 01, 2011 8:20 pm Post subject: Reply with quote
Here's a box plot of data by position:
EDIT: As per MikeG's suggestion below, the box plot (made using R) shows the median, 25th and 75th percentiles (ends of the box). I'm not quite clear on the whiskers. I initially thought they represented the 95th percentile, but according to the R documentation, the default is the following:
Quote:
the whiskers extend to the most extreme data point which is no more than range times the interquartile range from the box.
The parameter range is set to 1.5 as a default, which is what I used.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Last edited by EvanZ on Mon Jan 03, 2011 3:04 pm; edited 1 time in total
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 10:37 pm Post subject: Reply with quote
Ed Küpfer wrote:
Team offense is slightly more variable than defense, using a number of different metrics and methods. I'm not aware of anybody that has found anything different. As far as I know, no one has found out why.
Thanks, that's helpful. But if the difference is only slight, doesn't that suggest a problem with player metrics that seem to suggest far more variance of the offensive side? It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 10:46 pm Post subject: Reply with quote
EvanZ wrote:
Take two teams with the following ratings:
Code:
Team OFF DEF
A 105 100
B 100 95
If it's true that offense has more variance, Team B would win more, right?
No, I don't think that follows. It is more unusual to see a 95 DEF than a 105 OFF, but the win value is still identical. What should be true is that Team A will win more if:
A: OFF = +1 SD, DEF = Avg.
B: OFF = Avg, DEF = +1 SD
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sat Jan 01, 2011 10:56 pm Post subject: Reply with quote
Guy, here's a couple summary stats on OFF100 and DEF100:
Code:
STDEV MAD 1QUINT 3QUINT
OFF100 4.14 2.56 -0.61 2.71
DEF100 2.86 1.76 -2.74 -0.39
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin
Joined: 30 Dec 2004
Posts: 979
Location: Seattle
PostPosted: Sun Jan 02, 2011 12:36 am Post subject: Reply with quote
Guy wrote:
Thanks, that's helpful. But if the difference is only slight, doesn't that suggest a problem with player metrics that seem to suggest far more variance of the offensive side? It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Not to the extent that external factors -- specifically, coaches -- may influence the variance to different degrees on offense and defense. The adjusted plus-minus scores for coaches don't seem to be available anymore, but my memory is that they confirmed the conventional wisdom that coaches have a greater impact at the defensive end of the floor.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Sun Jan 02, 2011 2:02 am Post subject: Reply with quote
The simple average of the current 30 coaches was +0.5 pt positive impact on the defensive efficiency and -.13 pt negative impact on offensive efficiency.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC
PostPosted: Sun Jan 02, 2011 6:02 am Post subject: Reply with quote
Mike G wrote:
Many fans will have a problem with this language:
Quote:
... Dorell Wright was actually a negative contributer.
...
Do you really win 41 games with zero net contribution toward winning those many games?
The opening post solicited comments, and we soon veered off into tangents. But what about this question about the language describing EvanZ's results?
We all know that a zero point differential is average, and that half of all teams are outscored on a given night. This is not equivalent to saying that half of all player contributions are negative.
A team of players make (many) positive and (a few) negative contributions for 48 minutes. If your opponents make greater pos-negs, you lose. That doesn't negate your acts, it just outdoes them.
It's a misinterpretation of +/- data to state that below-average = negative contribution. DSMok1 suggests 'replacement level' as a viable true zero, and the challenge would be to find that level.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sun Jan 02, 2011 8:33 am Post subject: Reply with quote
Kevin Pelton wrote:
Guy wrote:
It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Not to the extent that external factors -- specifically, coaches -- may influence the variance to different degrees on offense and defense. The adjusted plus-minus scores for coaches don't seem to be available anymore, but my memory is that they confirmed the conventional wisdom that coaches have a greater impact at the defensive end of the floor.
Right, that could create a disparity in the respective relationships between player variance and team variance at each end. Very interesting. Still, it seems like that could only explain a fairly small portion of the difference that EvanZ is reporting here. If you break down APM into defense and offense, what are the SDs? Has someone done that? Presumably, APM wouldn't be impacted by any limitations of defensive statistics.
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sun Jan 02, 2011 9:38 am Post subject: Reply with quote
Mike G wrote:
Mike G wrote:
Many fans will have a problem with this language:
Quote:
... Dorell Wright was actually a negative contributer.
...
Do you really win 41 games with zero net contribution toward winning those many games?
The opening post solicited comments, and we soon veered off into tangents. But what about this question about the language describing EvanZ's results?
We all know that a zero point differential is average, and that half of all teams are outscored on a given night. This is not equivalent to saying that half of all player contributions are negative.
A team of players make (many) positive and (a few) negative contributions for 48 minutes. If your opponents make greater pos-negs, you lose. That doesn't negate your acts, it just outdoes them.
It's a misinterpretation of +/- data to state that below-average = negative contribution. DSMok1 suggests 'replacement level' as a viable true zero, and the challenge would be to find that level.
I admit my interpretation of "-" as negative is hard-wired into my brain. Semantics aside, I think VORP is useful. Guy suggested it on my post yesterday, too, and I responded that I would like to calculate a position-adjusted VORP using the first quintile as a reference. The first quintile for ezPM100 across all players is ~ -3.0 (-2.91 to be exact). This would make Vlad Rad a replacement quality player (even though he's making $6M or so). As you can see by the plot above, replacement level would vary by position.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC
PostPosted: Mon Jan 03, 2011 6:41 am Post subject: Reply with quote
EvanZ wrote:
...my interpretation of "-" as negative is hard-wired into my brain...
What if, instead of rating players by T-O (team minus opponent points), you rated them by T/(T+O) ?
This would be the % of all points scored by the player's team while he's on the floor.
Values would largely be in the .475-.525 range. A sort of personal winning %.
Another way would be simply T/O, such that 1.00 is the norm.
Code:
As you can see by the plot above, replacement level would vary by position.
Maybe you can edit that post with some explanation. I don't know what I'm seeing there.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 12:31 pm Post subject: Reply with quote
Guy wrote:
Evan: Is there a reason you think total offense and total defense shouldn't each add to zero?
It looks to me like the player variance on offense is much larger than on defense. Do you think this is just a limitation of the data that we have, or that offensive variance really is much larger?
After substituting parameters for the current season (PPP and rebounding rates), the offset on offense is now roughly +2.4 pt per game, and offset on defense is -1.8 pts/game. (EDIT: I originally calculated for 82 games, when only ~33 had been played. Whoops.) It appears that even slight approximation of the numbers (i.e. rounding to one decimal, instead of 2) is very significant, and was leading to the large offsets. At any rate, I've put up the data (from 12/31) on Google Docs as a csv:
https://spreadsheets.google.com/pub?key ... output=csv
Top 50
Code:
SEASON OVERALL ORANK DRANK LAST FIRST TEAM POS #POSS EZPM100 O100 D100
2010-2011 1 1 4 Paul Chris NOH 1 2295 9.26 6.21 3.06
2010-2011 2 13 11 Wade Dwyane MIA 2 2418 5.80 3.66 2.13
2010-2011 3 22 8 James LeBron MIA 3 2603 5.53 3.20 2.33
2010-2011 4 2 132 Nash Steve PHX 1 2094 5.24 5.69 -0.45
2010-2011 5 81 3 Howard Dwight ORL 5 2102 4.97 1.71 3.25
2010-2011 6 10 29 Ginobili Manu SAS 2 2278 4.90 3.83 1.07
2010-2011 7 3 112 Williams Deron UTA 1 2579 4.83 5.00 -0.16
2010-2011 8 109 1 Garnett Kevin BOS 4 1999 4.60 1.02 3.58
2010-2011 9 6 76 Gasol Pau LAL 4 2624 4.49 4.18 0.30
2010-2011 10 54 16 Rondo Rajon BOS 1 1649 4.15 2.40 1.75
2010-2011 11 37 28 Rose Derrick CHI 1 2321 4.09 2.90 1.18
2010-2011 12 9 96 Horford Al ATL 5 2246 3.98 3.97 0.01
2010-2011 13 40 32 Hill George SAS 1 1591 3.83 2.82 1.01
2010-2011 14 23 52 Parker Tony SAS 1 2264 3.83 3.20 0.63
2010-2011 15 20 62 Chandler Tyson DAL 5 1725 3.72 3.22 0.50
2010-2011 16 41 41 Miller Andre POR 1 2075 3.60 2.79 0.82
2010-2011 17 45 35 Carter Vince ORL 2 1312 3.55 2.62 0.93
2010-2011 18 4 174 Love Kevin MIN 4 2517 3.47 4.45 -0.97
2010-2011 19 5 168 Hilario Nene DEN 5 1824 3.47 4.38 -0.92
2010-2011 20 26 73 Noah Joakim CHI 5 1803 3.46 3.12 0.34
2010-2011 21 33 65 Brand Elton PHI 4 2066 3.43 2.95 0.49
2010-2011 22 92 14 Duncan Tim SAS 5 1997 3.39 1.41 1.98
2010-2011 23 82 18 Thomas Tyrus CHA 4 995 3.36 1.71 1.65
2010-2011 24 47 45 Barnes Matt LAL 3 1386 3.33 2.54 0.79
2010-2011 25 60 31 Brown Shannon LAL 2 1177 3.29 2.24 1.05
2010-2011 26 52 46 Pierce Paul BOS 3 2298 3.24 2.46 0.78
2010-2011 27 145 5 Daniels Marquis BOS 2 1138 3.23 0.48 2.76
2010-2011 28 31 82 Felton Raymond NYK 1 2599 3.22 3.03 0.18
2010-2011 29 108 10 Kidd Jason DAL 1 2099 3.21 1.04 2.17
2010-2011 30 21 101 Nowitzki Dirk DAL 4 2177 3.19 3.21 -0.02
2010-2011 31 48 55 Westbrook Russell OKC 1 2471 3.14 2.54 0.60
2010-2011 32 8 167 Johnson Amir TOR 4 1601 3.10 4.00 -0.89
2010-2011 33 56 47 Evans Reggie TOR 4 778 3.09 2.36 0.73
2010-2011 34 14 147 Nelson Jameer ORL 1 1655 2.97 3.60 -0.63
2010-2011 35 147 7 Brewer Ronnie CHI 3 1306 2.93 0.45 2.48
2010-2011 36 63 43 Boozer Carlos CHI 4 1057 2.90 2.09 0.81
2010-2011 37 24 124 Calderon Jose TOR 1 1526 2.88 3.18 -0.31
2010-2011 38 7 187 Martin Kevin HOU 2 2195 2.87 4.13 -1.26
2010-2011 39 97 19 Teague Jeff ATL 1 775 2.85 1.26 1.59
2010-2011 40 98 20 Camby Marcus POR 5 1643 2.83 1.25 1.58
2010-2011 41 27 122 Odom Lamar LAL 4 2434 2.82 3.12 -0.29
2010-2011 42 67 44 Lowry Kyle HOU 1 1969 2.80 2.01 0.80
2010-2011 43 12 173 Lawson Ty DEN 1 1724 2.74 3.71 -0.97
2010-2011 44 94 23 Gay Rudy MEM 3 2562 2.73 1.33 1.39
2010-2011 45 114 15 McGrady Tracy DET 3 1086 2.71 0.93 1.78
2010-2011 46 70 42 Anthony Carmelo DEN 3 1789 2.70 1.88 0.81
2010-2011 47 51 85 Bryant Kobe LAL 2 2232 2.63 2.46 0.16
2010-2011 48 19 146 McGee JaVale WAS 5 1561 2.62 3.24 -0.62
2010-2011 49 30 137 Randolph Zach MEM 4 2095 2.54 3.05 -0.50
2010-2011 50 17 162 Griffin Blake LAC 4 2447 2.53 3.32 -0.80
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Last edited by EvanZ on Wed Jan 05, 2011 4:14 pm; edited 1 time in total
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 1:33 pm Post subject: Reply with quote
Some (probably not many) here may be interested in how this compares to Wins Produced. I took the top 30 players according to WP and compared their rankings with ezPM100. Here they are in descending order of difference (i.e. more positive means ezPM100 rates them higher than WP):
Code:
Player...WP Rank...ezPM100 Rank...Delta
Deron Williams...24...7...17
Manu Ginobili...19...6...13
Dwyane Wade...10...2...8
Steve Nash...11...4...7
Pau Gasol...16...9...7
LeBron James...7...3...4
Nene Hilario...23...19...4
Tyson Chandler...17...15...2
Matt Barnes...26...24...2
Dwight Howard...6...5...1
Chris Paul...1...1...0
Rajon Rondo...9...10...-1
Paul Pierce...25...26...-1
Kevin Garnett...5...8...-3
Al Horford...8...12...-4
Ronnie Brewer...29...35...-6
Joakim Noah...13...20...-7
Carlos Boozer...28...36...-8
Jason Kidd...20...29...-9
Tim Duncan...12...22...-10
Kevin Love...2...18...-16
Lamar Odom...22...41...-19
Andrew Bogut...30...52...-22
JaVale McGee...21...48...-27
Blake Griffin...18...50...-32
Zach Randolph...15...49...-34
Marcus Camby...3...40...-37
Landry Fields...14...60...-46
Kris Humphries...4...58...-54
Josh Smith...27...89...-62
Sorry, it's not pretty.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Wed Jan 05, 2011 2:31 pm Post subject: Reply with quote
The 2 main distinctions / sources of rank movement between the other metric and yours that I see are your metric's division of credit between opponent missed shots and defensive rebounds and allocation of shot defense credit / blame to players based on play by play data when actually on the court.
Any other differences you think should be highlighted?
Looking at the movement up or down for names on the list and thinking about their defensive rebounding and shot defense, these components probably do a lot to explain the movement. But maybe not of all of it.
Blake Griffin's ranking is clearly impacted on your metric by its reduced share of credit going to defensive rebounds and rather being split among the defenders on the court and by a fairly weak team shot defense and very weak overall defense while he is on the court. Camby is hurt more by the rebounding change. By using team level shot defense though he probably isn't getting the boost that he could / perhaps "should" if his individual counterpart defense was recognized in any elevated way.
A few other player cases- particularly Williams and Nash- make me interested in identifying any other notable metric differences involved in the fairly large rank changes.
Last edited by Crow on Wed Jan 05, 2011 5:53 pm; edited 1 time in total
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 3:11 pm Post subject: Reply with quote
Crow wrote:
A few other player cases- particularly Williams and Nash- make me interested in identifying any other notable metric differences.
Nash surprised me, but he is having an incredible offensive season. 0.631 TS and the highest AST% (0.521) of his career.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Wed Jan 05, 2011 9:20 pm Post subject: Reply with quote
The metrics also differ notably on the cost of a missed shot (your metric charging less based on the rate of offensive rebounds) and the treatment of fouls given (your metric only charging for free throws taken) and these differences may also play a part in the rank change for Nash and Williams.
Different metrics, different weights, different rankings. Important to know and keep in mind why they are different.
What sort of testing you do intend to do? Predictive power and predictive power relative to other metrics is of course of interest and importance.
Metric stability and performance stability are topics bigger than I want to try to handle fully or quickly but I will say that the relative weights of the main components of player performance get in different metrics and their specific volatility will affect the overall metric volatility when using the results of one year to provide a simple carry-forward prediction of future years. Some modeling of year to year change at the component level or at the roll-up level based on patterns in the historical data may be "needed" or "helpful" to bolt on top of whatever basic player measurement model is used. Or step back from the focus on a point prediction of the future and view each year of data at a datapoint helping to define a true talent range and settle for that.
Author Message
gabefarkas
Joined: 31 Dec 2004
Posts: 1311
Location: Durham, NC
PostPosted: Thu Jan 06, 2011 3:24 pm Post subject: Reply with quote
EvanZ wrote:
deepak wrote:
EvanZ wrote:
BTW, how do you get the formatting so nice? Not by hand, I assume.
I copied it into Excel (text to columns), merged the first and last name columns. Then copied it into my text editor (vim), replaced tabs with spaces, and adjusted the spacing between the columns (easy to do in vim).
ah, almost by hand. Laughing
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
Back to top
View user's profile Send private message Send e-mail AIM Address
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Thu Jan 06, 2011 3:44 pm Post subject: Reply with quote
gabefarkas wrote:
ah, almost by hand. Laughing
Search for a program called "Tabs2Spaces", it does a lot of that work for you.[/quote]
I'm a Mac guy! Crying or Very sad
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 555
Location: Where the wind comes sweeping down the plains
PostPosted: Thu Jan 06, 2011 3:59 pm Post subject: Reply with quote
EvanZ wrote:
gabefarkas wrote:
ah, almost by hand. :lol:Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
It's not that hard to write a script for this EvanZ, really it's not. That or you should just get Excel and use the VBA I wrote!
_________________
GodismyJudgeOK.com/DStats
Back to top
View user's profile Send private message Visit poster's website
gabefarkas
Joined: 31 Dec 2004
Posts: 1311
Location: Durham, NC
PostPosted: Sat Jan 08, 2011 2:50 pm Post subject: Reply with quote
EvanZ wrote:
gabefarkas wrote:
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
That's too bad. Laughing
Back to top
View user's profile Send private message Send e-mail AIM Address
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Sat Jan 08, 2011 4:26 pm Post subject: Reply with quote
gabefarkas wrote:
EvanZ wrote:
gabefarkas wrote:
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
That's too bad. Laughing
Believe me, after years of Windows hell, I don't miss it.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Mon Jan 10, 2011 9:09 pm Post subject: top rookies (1/08) Reply with quote
Top Rookies (>250 possessions)
Code:
RANK NAME TEAM POSSESSIONS ezPM100
1 Landry Fields NYK 2467 2.47
2 Blake Griffin LAC 2629 2.34
3 Jeremy Lin GSW 256 1.14
4 Ed Davis TOR 745 0.68
5 Omer Asik CHI 850 0.32
6 John Wall WAS 1515 0.28
7 Derrick Favors NJN 1159 0.19
8 Paul George IND 276 0.03
9 Gary Neal SAS 1271 -0.86
10 Eric Bledsoe LAC 1699 -0.94
11 Al-Farouq Aminu LAC 1152 -0.97
12 Gary Forbes DEN 849 -1.06
13 Eugene Jeter SAC 545 -1.32
14 Larry Sanders MIL 630 -1.43
15 Ekpe Udoh GSW 328 -1.51
16 Tiago Splitter SAS 577 -1.52
17 Greg Monroe DET 1124 -1.61
18 Ishmael Smith HOU 588 -1.65
19 Trevor Booker WAS 609 -2.05
20 Quincy Pondexter NOH 472 -2.17
21 Wesley Johnson MIN 2236 -2.62
22 Gordon Hayward UTA 693 -2.66
23 Greivis Vasquez MEM 732 -2.95
24 Evan Turner PHI 1567 -2.96
25 Xavier Henry MEM 889 -3.40
26 Armon Johnson POR 384 -3.68
27 DeMarcus Cousins SAC 1591 -4.32
28 Semih Erden BOS 576 -4.93
29 Nikola Pekovic MIN 580 -5.98
30 Timofey Mozgov NYK 481 -7.50
31 Manny Harris CLE 312 -7.70
32 Derrick Caracter LAL 268 -8.36
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sat Jan 01, 2011 8:20 pm Post subject: Reply with quote
Here's a box plot of data by position:
EDIT: As per MikeG's suggestion below, the box plot (made using R) shows the median, 25th and 75th percentiles (ends of the box). I'm not quite clear on the whiskers. I initially thought they represented the 95th percentile, but according to the R documentation, the default is the following:
Quote:
the whiskers extend to the most extreme data point which is no more than range times the interquartile range from the box.
The parameter range is set to 1.5 as a default, which is what I used.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Last edited by EvanZ on Mon Jan 03, 2011 3:04 pm; edited 1 time in total
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 10:37 pm Post subject: Reply with quote
Ed Küpfer wrote:
Team offense is slightly more variable than defense, using a number of different metrics and methods. I'm not aware of anybody that has found anything different. As far as I know, no one has found out why.
Thanks, that's helpful. But if the difference is only slight, doesn't that suggest a problem with player metrics that seem to suggest far more variance of the offensive side? It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Back to top
View user's profile Send private message
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sat Jan 01, 2011 10:46 pm Post subject: Reply with quote
EvanZ wrote:
Take two teams with the following ratings:
Code:
Team OFF DEF
A 105 100
B 100 95
If it's true that offense has more variance, Team B would win more, right?
No, I don't think that follows. It is more unusual to see a 95 DEF than a 105 OFF, but the win value is still identical. What should be true is that Team A will win more if:
A: OFF = +1 SD, DEF = Avg.
B: OFF = Avg, DEF = +1 SD
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sat Jan 01, 2011 10:56 pm Post subject: Reply with quote
Guy, here's a couple summary stats on OFF100 and DEF100:
Code:
STDEV MAD 1QUINT 3QUINT
OFF100 4.14 2.56 -0.61 2.71
DEF100 2.86 1.76 -2.74 -0.39
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin
Joined: 30 Dec 2004
Posts: 979
Location: Seattle
PostPosted: Sun Jan 02, 2011 12:36 am Post subject: Reply with quote
Guy wrote:
Thanks, that's helpful. But if the difference is only slight, doesn't that suggest a problem with player metrics that seem to suggest far more variance of the offensive side? It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Not to the extent that external factors -- specifically, coaches -- may influence the variance to different degrees on offense and defense. The adjusted plus-minus scores for coaches don't seem to be available anymore, but my memory is that they confirmed the conventional wisdom that coaches have a greater impact at the defensive end of the floor.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Sun Jan 02, 2011 2:02 am Post subject: Reply with quote
The simple average of the current 30 coaches was +0.5 pt positive impact on the defensive efficiency and -.13 pt negative impact on offensive efficiency.
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC
PostPosted: Sun Jan 02, 2011 6:02 am Post subject: Reply with quote
Mike G wrote:
Many fans will have a problem with this language:
Quote:
... Dorell Wright was actually a negative contributer.
...
Do you really win 41 games with zero net contribution toward winning those many games?
The opening post solicited comments, and we soon veered off into tangents. But what about this question about the language describing EvanZ's results?
We all know that a zero point differential is average, and that half of all teams are outscored on a given night. This is not equivalent to saying that half of all player contributions are negative.
A team of players make (many) positive and (a few) negative contributions for 48 minutes. If your opponents make greater pos-negs, you lose. That doesn't negate your acts, it just outdoes them.
It's a misinterpretation of +/- data to state that below-average = negative contribution. DSMok1 suggests 'replacement level' as a viable true zero, and the challenge would be to find that level.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Guy
Joined: 02 May 2007
Posts: 128
PostPosted: Sun Jan 02, 2011 8:33 am Post subject: Reply with quote
Kevin Pelton wrote:
Guy wrote:
It seems to me the team variance is telling us that the variances in defensive production and offensive production should be similar.
Not to the extent that external factors -- specifically, coaches -- may influence the variance to different degrees on offense and defense. The adjusted plus-minus scores for coaches don't seem to be available anymore, but my memory is that they confirmed the conventional wisdom that coaches have a greater impact at the defensive end of the floor.
Right, that could create a disparity in the respective relationships between player variance and team variance at each end. Very interesting. Still, it seems like that could only explain a fairly small portion of the difference that EvanZ is reporting here. If you break down APM into defense and offense, what are the SDs? Has someone done that? Presumably, APM wouldn't be impacted by any limitations of defensive statistics.
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Sun Jan 02, 2011 9:38 am Post subject: Reply with quote
Mike G wrote:
Mike G wrote:
Many fans will have a problem with this language:
Quote:
... Dorell Wright was actually a negative contributer.
...
Do you really win 41 games with zero net contribution toward winning those many games?
The opening post solicited comments, and we soon veered off into tangents. But what about this question about the language describing EvanZ's results?
We all know that a zero point differential is average, and that half of all teams are outscored on a given night. This is not equivalent to saying that half of all player contributions are negative.
A team of players make (many) positive and (a few) negative contributions for 48 minutes. If your opponents make greater pos-negs, you lose. That doesn't negate your acts, it just outdoes them.
It's a misinterpretation of +/- data to state that below-average = negative contribution. DSMok1 suggests 'replacement level' as a viable true zero, and the challenge would be to find that level.
I admit my interpretation of "-" as negative is hard-wired into my brain. Semantics aside, I think VORP is useful. Guy suggested it on my post yesterday, too, and I responded that I would like to calculate a position-adjusted VORP using the first quintile as a reference. The first quintile for ezPM100 across all players is ~ -3.0 (-2.91 to be exact). This would make Vlad Rad a replacement quality player (even though he's making $6M or so). As you can see by the plot above, replacement level would vary by position.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G
Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC
PostPosted: Mon Jan 03, 2011 6:41 am Post subject: Reply with quote
EvanZ wrote:
...my interpretation of "-" as negative is hard-wired into my brain...
What if, instead of rating players by T-O (team minus opponent points), you rated them by T/(T+O) ?
This would be the % of all points scored by the player's team while he's on the floor.
Values would largely be in the .475-.525 range. A sort of personal winning %.
Another way would be simply T/O, such that 1.00 is the norm.
Code:
As you can see by the plot above, replacement level would vary by position.
Maybe you can edit that post with some explanation. I don't know what I'm seeing there.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 12:31 pm Post subject: Reply with quote
Guy wrote:
Evan: Is there a reason you think total offense and total defense shouldn't each add to zero?
It looks to me like the player variance on offense is much larger than on defense. Do you think this is just a limitation of the data that we have, or that offensive variance really is much larger?
After substituting parameters for the current season (PPP and rebounding rates), the offset on offense is now roughly +2.4 pt per game, and offset on defense is -1.8 pts/game. (EDIT: I originally calculated for 82 games, when only ~33 had been played. Whoops.) It appears that even slight approximation of the numbers (i.e. rounding to one decimal, instead of 2) is very significant, and was leading to the large offsets. At any rate, I've put up the data (from 12/31) on Google Docs as a csv:
https://spreadsheets.google.com/pub?key ... output=csv
Top 50
Code:
SEASON OVERALL ORANK DRANK LAST FIRST TEAM POS #POSS EZPM100 O100 D100
2010-2011 1 1 4 Paul Chris NOH 1 2295 9.26 6.21 3.06
2010-2011 2 13 11 Wade Dwyane MIA 2 2418 5.80 3.66 2.13
2010-2011 3 22 8 James LeBron MIA 3 2603 5.53 3.20 2.33
2010-2011 4 2 132 Nash Steve PHX 1 2094 5.24 5.69 -0.45
2010-2011 5 81 3 Howard Dwight ORL 5 2102 4.97 1.71 3.25
2010-2011 6 10 29 Ginobili Manu SAS 2 2278 4.90 3.83 1.07
2010-2011 7 3 112 Williams Deron UTA 1 2579 4.83 5.00 -0.16
2010-2011 8 109 1 Garnett Kevin BOS 4 1999 4.60 1.02 3.58
2010-2011 9 6 76 Gasol Pau LAL 4 2624 4.49 4.18 0.30
2010-2011 10 54 16 Rondo Rajon BOS 1 1649 4.15 2.40 1.75
2010-2011 11 37 28 Rose Derrick CHI 1 2321 4.09 2.90 1.18
2010-2011 12 9 96 Horford Al ATL 5 2246 3.98 3.97 0.01
2010-2011 13 40 32 Hill George SAS 1 1591 3.83 2.82 1.01
2010-2011 14 23 52 Parker Tony SAS 1 2264 3.83 3.20 0.63
2010-2011 15 20 62 Chandler Tyson DAL 5 1725 3.72 3.22 0.50
2010-2011 16 41 41 Miller Andre POR 1 2075 3.60 2.79 0.82
2010-2011 17 45 35 Carter Vince ORL 2 1312 3.55 2.62 0.93
2010-2011 18 4 174 Love Kevin MIN 4 2517 3.47 4.45 -0.97
2010-2011 19 5 168 Hilario Nene DEN 5 1824 3.47 4.38 -0.92
2010-2011 20 26 73 Noah Joakim CHI 5 1803 3.46 3.12 0.34
2010-2011 21 33 65 Brand Elton PHI 4 2066 3.43 2.95 0.49
2010-2011 22 92 14 Duncan Tim SAS 5 1997 3.39 1.41 1.98
2010-2011 23 82 18 Thomas Tyrus CHA 4 995 3.36 1.71 1.65
2010-2011 24 47 45 Barnes Matt LAL 3 1386 3.33 2.54 0.79
2010-2011 25 60 31 Brown Shannon LAL 2 1177 3.29 2.24 1.05
2010-2011 26 52 46 Pierce Paul BOS 3 2298 3.24 2.46 0.78
2010-2011 27 145 5 Daniels Marquis BOS 2 1138 3.23 0.48 2.76
2010-2011 28 31 82 Felton Raymond NYK 1 2599 3.22 3.03 0.18
2010-2011 29 108 10 Kidd Jason DAL 1 2099 3.21 1.04 2.17
2010-2011 30 21 101 Nowitzki Dirk DAL 4 2177 3.19 3.21 -0.02
2010-2011 31 48 55 Westbrook Russell OKC 1 2471 3.14 2.54 0.60
2010-2011 32 8 167 Johnson Amir TOR 4 1601 3.10 4.00 -0.89
2010-2011 33 56 47 Evans Reggie TOR 4 778 3.09 2.36 0.73
2010-2011 34 14 147 Nelson Jameer ORL 1 1655 2.97 3.60 -0.63
2010-2011 35 147 7 Brewer Ronnie CHI 3 1306 2.93 0.45 2.48
2010-2011 36 63 43 Boozer Carlos CHI 4 1057 2.90 2.09 0.81
2010-2011 37 24 124 Calderon Jose TOR 1 1526 2.88 3.18 -0.31
2010-2011 38 7 187 Martin Kevin HOU 2 2195 2.87 4.13 -1.26
2010-2011 39 97 19 Teague Jeff ATL 1 775 2.85 1.26 1.59
2010-2011 40 98 20 Camby Marcus POR 5 1643 2.83 1.25 1.58
2010-2011 41 27 122 Odom Lamar LAL 4 2434 2.82 3.12 -0.29
2010-2011 42 67 44 Lowry Kyle HOU 1 1969 2.80 2.01 0.80
2010-2011 43 12 173 Lawson Ty DEN 1 1724 2.74 3.71 -0.97
2010-2011 44 94 23 Gay Rudy MEM 3 2562 2.73 1.33 1.39
2010-2011 45 114 15 McGrady Tracy DET 3 1086 2.71 0.93 1.78
2010-2011 46 70 42 Anthony Carmelo DEN 3 1789 2.70 1.88 0.81
2010-2011 47 51 85 Bryant Kobe LAL 2 2232 2.63 2.46 0.16
2010-2011 48 19 146 McGee JaVale WAS 5 1561 2.62 3.24 -0.62
2010-2011 49 30 137 Randolph Zach MEM 4 2095 2.54 3.05 -0.50
2010-2011 50 17 162 Griffin Blake LAC 4 2447 2.53 3.32 -0.80
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Last edited by EvanZ on Wed Jan 05, 2011 4:14 pm; edited 1 time in total
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 1:33 pm Post subject: Reply with quote
Some (probably not many) here may be interested in how this compares to Wins Produced. I took the top 30 players according to WP and compared their rankings with ezPM100. Here they are in descending order of difference (i.e. more positive means ezPM100 rates them higher than WP):
Code:
Player...WP Rank...ezPM100 Rank...Delta
Deron Williams...24...7...17
Manu Ginobili...19...6...13
Dwyane Wade...10...2...8
Steve Nash...11...4...7
Pau Gasol...16...9...7
LeBron James...7...3...4
Nene Hilario...23...19...4
Tyson Chandler...17...15...2
Matt Barnes...26...24...2
Dwight Howard...6...5...1
Chris Paul...1...1...0
Rajon Rondo...9...10...-1
Paul Pierce...25...26...-1
Kevin Garnett...5...8...-3
Al Horford...8...12...-4
Ronnie Brewer...29...35...-6
Joakim Noah...13...20...-7
Carlos Boozer...28...36...-8
Jason Kidd...20...29...-9
Tim Duncan...12...22...-10
Kevin Love...2...18...-16
Lamar Odom...22...41...-19
Andrew Bogut...30...52...-22
JaVale McGee...21...48...-27
Blake Griffin...18...50...-32
Zach Randolph...15...49...-34
Marcus Camby...3...40...-37
Landry Fields...14...60...-46
Kris Humphries...4...58...-54
Josh Smith...27...89...-62
Sorry, it's not pretty.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Wed Jan 05, 2011 2:31 pm Post subject: Reply with quote
The 2 main distinctions / sources of rank movement between the other metric and yours that I see are your metric's division of credit between opponent missed shots and defensive rebounds and allocation of shot defense credit / blame to players based on play by play data when actually on the court.
Any other differences you think should be highlighted?
Looking at the movement up or down for names on the list and thinking about their defensive rebounding and shot defense, these components probably do a lot to explain the movement. But maybe not of all of it.
Blake Griffin's ranking is clearly impacted on your metric by its reduced share of credit going to defensive rebounds and rather being split among the defenders on the court and by a fairly weak team shot defense and very weak overall defense while he is on the court. Camby is hurt more by the rebounding change. By using team level shot defense though he probably isn't getting the boost that he could / perhaps "should" if his individual counterpart defense was recognized in any elevated way.
A few other player cases- particularly Williams and Nash- make me interested in identifying any other notable metric differences involved in the fairly large rank changes.
Last edited by Crow on Wed Jan 05, 2011 5:53 pm; edited 1 time in total
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 293
PostPosted: Wed Jan 05, 2011 3:11 pm Post subject: Reply with quote
Crow wrote:
A few other player cases- particularly Williams and Nash- make me interested in identifying any other notable metric differences.
Nash surprised me, but he is having an incredible offensive season. 0.631 TS and the highest AST% (0.521) of his career.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Crow
Joined: 20 Jan 2009
Posts: 817
PostPosted: Wed Jan 05, 2011 9:20 pm Post subject: Reply with quote
The metrics also differ notably on the cost of a missed shot (your metric charging less based on the rate of offensive rebounds) and the treatment of fouls given (your metric only charging for free throws taken) and these differences may also play a part in the rank change for Nash and Williams.
Different metrics, different weights, different rankings. Important to know and keep in mind why they are different.
What sort of testing you do intend to do? Predictive power and predictive power relative to other metrics is of course of interest and importance.
Metric stability and performance stability are topics bigger than I want to try to handle fully or quickly but I will say that the relative weights of the main components of player performance get in different metrics and their specific volatility will affect the overall metric volatility when using the results of one year to provide a simple carry-forward prediction of future years. Some modeling of year to year change at the component level or at the roll-up level based on patterns in the historical data may be "needed" or "helpful" to bolt on top of whatever basic player measurement model is used. Or step back from the focus on a point prediction of the future and view each year of data at a datapoint helping to define a true talent range and settle for that.
Author Message
gabefarkas
Joined: 31 Dec 2004
Posts: 1311
Location: Durham, NC
PostPosted: Thu Jan 06, 2011 3:24 pm Post subject: Reply with quote
EvanZ wrote:
deepak wrote:
EvanZ wrote:
BTW, how do you get the formatting so nice? Not by hand, I assume.
I copied it into Excel (text to columns), merged the first and last name columns. Then copied it into my text editor (vim), replaced tabs with spaces, and adjusted the spacing between the columns (easy to do in vim).
ah, almost by hand. Laughing
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
Back to top
View user's profile Send private message Send e-mail AIM Address
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Thu Jan 06, 2011 3:44 pm Post subject: Reply with quote
gabefarkas wrote:
ah, almost by hand. Laughing
Search for a program called "Tabs2Spaces", it does a lot of that work for you.[/quote]
I'm a Mac guy! Crying or Very sad
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
DSMok1
Joined: 05 Aug 2009
Posts: 555
Location: Where the wind comes sweeping down the plains
PostPosted: Thu Jan 06, 2011 3:59 pm Post subject: Reply with quote
EvanZ wrote:
gabefarkas wrote:
ah, almost by hand. :lol:Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
It's not that hard to write a script for this EvanZ, really it's not. That or you should just get Excel and use the VBA I wrote!
_________________
GodismyJudgeOK.com/DStats
Back to top
View user's profile Send private message Visit poster's website
gabefarkas
Joined: 31 Dec 2004
Posts: 1311
Location: Durham, NC
PostPosted: Sat Jan 08, 2011 2:50 pm Post subject: Reply with quote
EvanZ wrote:
gabefarkas wrote:
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
That's too bad. Laughing
Back to top
View user's profile Send private message Send e-mail AIM Address
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Sat Jan 08, 2011 4:26 pm Post subject: Reply with quote
gabefarkas wrote:
EvanZ wrote:
gabefarkas wrote:
Search for a program called "Tabs2Spaces", it does a lot of that work for you.
I'm a Mac guy! Crying or Very sad
That's too bad. Laughing
Believe me, after years of Windows hell, I don't miss it.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
EvanZ
Joined: 22 Nov 2010
Posts: 203
PostPosted: Mon Jan 10, 2011 9:09 pm Post subject: top rookies (1/08) Reply with quote
Top Rookies (>250 possessions)
Code:
RANK NAME TEAM POSSESSIONS ezPM100
1 Landry Fields NYK 2467 2.47
2 Blake Griffin LAC 2629 2.34
3 Jeremy Lin GSW 256 1.14
4 Ed Davis TOR 745 0.68
5 Omer Asik CHI 850 0.32
6 John Wall WAS 1515 0.28
7 Derrick Favors NJN 1159 0.19
8 Paul George IND 276 0.03
9 Gary Neal SAS 1271 -0.86
10 Eric Bledsoe LAC 1699 -0.94
11 Al-Farouq Aminu LAC 1152 -0.97
12 Gary Forbes DEN 849 -1.06
13 Eugene Jeter SAC 545 -1.32
14 Larry Sanders MIL 630 -1.43
15 Ekpe Udoh GSW 328 -1.51
16 Tiago Splitter SAS 577 -1.52
17 Greg Monroe DET 1124 -1.61
18 Ishmael Smith HOU 588 -1.65
19 Trevor Booker WAS 609 -2.05
20 Quincy Pondexter NOH 472 -2.17
21 Wesley Johnson MIN 2236 -2.62
22 Gordon Hayward UTA 693 -2.66
23 Greivis Vasquez MEM 732 -2.95
24 Evan Turner PHI 1567 -2.96
25 Xavier Henry MEM 889 -3.40
26 Armon Johnson POR 384 -3.68
27 DeMarcus Cousins SAC 1591 -4.32
28 Semih Erden BOS 576 -4.93
29 Nikola Pekovic MIN 580 -5.98
30 Timofey Mozgov NYK 481 -7.50
31 Manny Harris CLE 312 -7.70
32 Derrick Caracter LAL 268 -8.36
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir