This is Google's cache of viewtopic.php?t=33. It is a snapshot of the page as it appeared on Jan 22, 2011 15:42:45 GMT. The current page could have changed in the meantime. Learn more

Text-only version
These search terms are highlighted: ed küpfer  
APBRmetrics :: View topic - Expected Win-Loss differences?
APBRmetrics Forum Index APBRmetrics
The statistical revolution will not be televised.
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Expected Win-Loss differences?

 
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion
View previous topic :: View next topic  
Author Message
Nikos



Joined: 16 Jan 2005
Posts: 346

PostPosted: Wed Jan 19, 2005 2:17 am    Post subject: Expected Win-Loss differences? Reply with quote

Why are B-R.com and BR.com's Expected Win%'s for a team reasonably different in results? Which is more accurate?
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 763
Location: Toronto

PostPosted: Wed Jan 19, 2005 3:17 am    Post subject: Re: Expected Win-Loss differences? Reply with quote

Nikos wrote:
Why are B-R.com and BR.com's Expected Win%'s for a team reasonably different in results? Which is more accurate?

B-R's is. But the differences are small:

Both esitmates use the Pythagorean Method. B-R uses an exponent of 14, giving a RMSE of 3.36. BR uses 16.5, with a RMSE of 3.81. To give you an idea of what those differences mean in less abstract terms, the 14 exponent gave an answer within 1 game of the actual win total 35.7% of the time, while the 16.5 exponent worked on only 32.2% of the time -- which works out, roughly, to 1 team more per season.

Code:
Error        Pyth14  Pyth16.5
+/- 1 game    35.7%   32.2%
+/- 2 games   56.5%   50.3%

RMSE           3.36    3.81

BTW if you're not familiar with the RMSE (Root Mean Squared Error), it is a meaure of the amount of variation of the errors. Roughly 2/3 of the errors will fall within 1 RMSE of zero in this case.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
Nikos



Joined: 16 Jan 2005
Posts: 346

PostPosted: Wed Jan 19, 2005 3:19 am    Post subject: Reply with quote

Interesting, I noticed B-R.com has Expected W-L's to be 2-3 games less usually than BR.com's. I noticed it for the Spurs the past few seasons, they are always 2-3 games less.

And no I'm not familiar with your analysis below etc... I'll try and look into a bit, and learn Smile
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 763
Location: Toronto

PostPosted: Wed Jan 19, 2005 3:57 am    Post subject: Reply with quote

I'll go into a bit more detail.

Both sites use the Pythagorean Method of estimating WIN%. It goes like this:
Code:
WIN% =               Points Scored^x
           ----------------------------------
           Points Scored^x + Points Allowed^x

where x an exponent, usually somewhere between 10 and 20. B-R.com has used 14 as their exponent, BR.com has used 16.5.

For example, last season the Pistons scored 7388 points, and allowed 6909. B-R would calculate their expected WIN% like this:
Code:
WIN% = 7388^14 / (7388^14 + 6909^14)
     = 71.9%
     = 58.9 wins


BR would do it like this:
Code:
WIN% = 7388^16.5 / (7388^16.5 + 6909^16.5)
     = 75.1%
     = 61.6 wins

Detroit actually won 54 games, so B-R was off by 4.9 games, and BR was off by 7.6.

If you look at the whole league, it looks like this:
Code:
TEAM    WINS   B-R.com Error  BR.com  Error
ATL     28      27.5   -0.5    25.3   -2.2
BOS     36      36.7    0.7    35.9   -0.8
CHI     23      22.7   -0.3    20.0   -2.7
CLE     35      33.1   -1.9    31.7   -1.4
DAL     52      53.0    1.0    55.0    2.0
DEN     43      44.2    1.2    44.7    0.6
DET     54      58.9    4.9    61.6    2.7
GS      37      38.8    1.8    38.4   -0.4
HOU     45      46.6    1.6    47.5    1.0
IND     61      58.5   -2.5    61.1    2.6
LAC     28      27.9   -0.1    25.8   -2.1
LAK     56      52.3   -3.7    54.2    1.9
MEM     50      48.3   -1.7    49.5    1.3
MIA     42      42.7    0.7    43.0    0.3
MIL     41      44.1    3.1    44.7    0.6
MIN     58      57.2   -0.8    59.7    2.5
NJ      47      49.0    2.0    50.3    1.4
NO      41      40.7   -0.3    40.6   -0.1
NY      39      36.5   -2.5    35.6   -0.8
ORL     21      21.9    0.9    19.1   -2.8
PHI     33      33.1    0.1    31.7   -1.4
PHX     29      30.1    1.1    28.3   -1.8
POR     41      37.0   -4.0    36.3   -0.7
SAC     55      54.8   -0.2    57.0    2.2
SAN     57      62.3    5.3    65.2    2.9
SEA     37      39.1    2.1    38.8   -0.3
TRN     33      31.2   -1.8    29.6   -1.7
UTA     42      37.1   -4.9    36.4   -0.7
WAZ     25      24.8   -0.2    22.3   -2.5


You can see that 12 of B-R's estimates (41%) fell within +/- 1 game of the team's actual win total. 11 (38%) of BR's estimates were within 1 game. 20 (69%) of B-R's estimates were within 2 games, while that number was 19 (66%) for BR.

2004 was actually one of the better seasons for Pythagorean predictions, both B-R's and BR's, but if you look through the history of the NBA, you can see a pattern repeated: B-R estimates are slightly more accurate than BR's. and both will predict the actual win total within 2 games of the true win total a little more than half the time.

Hope that helps.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
jkubatko



Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH

PostPosted: Wed Jan 19, 2005 12:35 pm    Post subject: Reply with quote

Nice summary Ed. Thanks for saving me the time.
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Display posts from previous:   
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion All times are GMT - 5 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group