[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Gnubg - Snowie error rate comparison

From: jorma kala
Subject: Re: [Bug-gnubg] Gnubg - Snowie error rate comparison
Date: Tue, 13 Nov 2007 13:25:42 +0100

how is the total error rate and the per move error rate exactly calculated?
Is it just the average difference in MWC of each unforced move with respect to the MWC of the optimal move?

On 11/10/07, Christian Anthon <address@hidden> wrote:
Snowie and GNU backgammon calculate error rate differently

From an example game:

Chequer Play Statistics:
Total moves                                           26
Unforced moves                                      22

Cube Statistics:
Total cube decisions                                   4
Close or actual cube decisions                         3

Error rate (total)                                  -0.145
Error rate (per decision)                          -5.8
Equiv. Snowie error rate                          -2.8

gnubg error = 0.145/(22+3)*1000 = 5.8
snowie error = 0.145/(26*2)*1000 = 2.8

So as you can see GNU backgammon attempts to use the number of actual
decisions, while Snowie only uses the number of moves. In my
experience the factor between the two error rates is closer a factor
of 2.


Bug-gnubg mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]