[EM] Which Condorcet completion method is best?

[Markus Schulze]

Dear Rob,

in the following table, I have added the statistical error.
Suppose that an experiment is positive P times and negative
N times, then the standard deviation is sqrt((P*N)/(P+N)).


Times method chooses the ratings winner
-----------------------------------------------------------------
 4 candidates           10 candidates          25 candidates
 BOR  79372 +/- 128     BOR  80474 +/- 125     BOR  83738 +/- 117
 BPm  73842 +/- 139     BPm  67741 +/- 148     BPm  63216 +/- 152
 TIm  73826 +/- 139     COm  67504 +/- 148     COm  62946 +/- 153
 COm  73814 +/- 139     TIm  66866 +/- 149     TIm  61195 +/- 154
 TIt  73373 +/- 140     BPw  65710 +/- 150     BPw  59852 +/- 155
 TIw  73183 +/- 140     TIt  65639 +/- 150     BPt  59729 +/- 155
 BPw  73166 +/- 140     BPt  65595 +/- 150     COw  59577 +/- 155
 COw  73125 +/- 140     COw  65450 +/- 150     TIt  59472 +/- 155
 BPt  73061 +/- 140     TIw  65441 +/- 150     COt  59465 +/- 155
 COt  73042 +/- 140     COt  65346 +/- 151     TIw  59320 +/- 155
 BPa  72838 +/- 141     BPa  64378 +/- 151     COP  58612 +/- 156
 COa  72016 +/- 142     COa  63184 +/- 153     BPa  57812 +/- 156
 COP  68766 +/- 147     COP  59883 +/- 155     COa  56907 +/- 157


In many cases the difference between two methods is below or
around the expected statistical error.

Markus Schulze