[Bug-gnubg] error rate and match winning chances

[Mueller Achim]

Hi folks,

I have a question regarding the error rates and match winning chances.

                                                                        A       
                                B
Overall Statistics:     
Error total EMG (MWC)                   -3.935 (-28.802%)       -4.498 
(-25.202%)
Error rate mEMG (MWC)                   -8.3   ( -0.061%)               -9.6   
( -0.054%)
Snowie error rate                                       -4.0   ( +0.000%)       
        -4.6   ( +0.000%)
Overall rating                                  Advanced                        
Advanced
Actual result                                           -50.00%                 
        +50.00%
Luck adjusted result                            -17.57%                         
+17.57%
Luck based FIBS rating diff.            -139.10
Error based abs. FIBS rating            1966.7                          1955.2
Chequerplay errors rating loss          69.9                            73.1
Cube errors rating loss                 13.4                            21.6


Luck Statistics:
Luck total EMG (MWC)                    -5.700 (-12.144%)       -2.564 
(+20.291%)
Luck rate mEMG (MWC)                    -11.5   ( -0.025%)              -5.2   
( +0.041%)
Luck rating                                             None                    
        None



Some Snowie numbers (same match to 21 point):

Error rate:                                     3.377           4.441
Match winning chance:           53.27%          46.73%
Average luck rate:                      -3.669          3.669


using gnubg's Luck adjusted result player B was 67.57% favorite and up  
139.10 fibs rating point?! Comparing all these numbers with th Snowie  
analysis I'm totally puzzled now. Also see this thread from may: http://www.nabble.com/match-winning-chances-td17190905.html 
 (which I tried to follow, but ...).

Which gnubg numbers are the most reliable for computing the match  
winning chances in a match to a certain amount of points (based on the  
error rates)? Is there a formula which can be used for computing match  
winning chances for other match lengths based on the given numbers of  
a analysis, e.g. take this 21p match and compute the chances in a 13p  
match.

Ciao

Achim