Hi, As the original poster in this thread, I want to thank all who contributed insights. It is a good reminder that statistic results are often not obvious/expected (mostly because the model one assumes is too simple and reality has always some surprises for you).
I ran similar rounds of tests with pachi and observed also a bias. In that case, it was because I wasn't using --pass_all_alive which resulted in games that were not really finished and the score evaluation seemed to have a bias in this case. After adding that parameter, the results were much closer to what I expected. (I ran several series of over 1000 games and never saw the kind of bias that was there before) It is possible that there was a similar thing with fuego. Or an "unlucky" streak. best regards, Vlad _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
