Author: blass uri
Date: 13:55:25 12/15/98
Go up one level in this thread
On December 15, 1998 at 15:12:07, Dan Kiski wrote: >On December 15, 1998 at 14:08:56, blass uri wrote: > >> >>On December 15, 1998 at 13:52:29, blass uri wrote: >> >>> >>>On December 15, 1998 at 04:52:19, Dan Kiski wrote: >>> >>>> >>>>As stated a few days ago when the subject of CM6555 and CM Faber\Pilz came along >>>>I had found Faber\Pilz beat the standard CM at all time controls. >>>> >>>>Harold Faber asked if I had tested at 2h\40 I stated I had but would re-test. >>>> >>>>First to state settings. CM 6555 taken from >>>>http://www.konts.lv/usr/Didzis/index.html opening book as stated. >>>> >>>>I am still not sure that CM6555 is any different from faber\Pilz. >>>> >>>>All games played on two identical P233 MMX machines each with 64 meg ram. >>>>CM hash at 32 meg. >>>> >>>>Time controls 40 moves in 120 minutes, balance 30 minutes. >>>> >>>> W L D GAMES SCORE >>>>CM 6555 72 52 76 200 110 >>>>CM 6000 52 72 76 200 90 >>> >>>The result is not clearly significant >>> >>>If you give 2 equal programs to play 200 games and if you assume that the >>>probability for white to win is 40% and that the probability for black to win is >>>30% >>>then the probability of a result of at least 110:90 for one side is close to 10% >>> >>>The probability of a result at least 110:90 for chessmaster 6555 is close to 5% >>>assuming that the programs are equal and the same for a result of at least >>>110:90 for chessmaster6000. > >I must be missing something since each engine had 100 whites and 100 blacks This is exactly my assumption then >I don't see your point. The probability for white to win is relevant to decide (for example in the extreme case when the probability for white to win is 100% between equal programs then I can be sure that the result between equal programs will be 100:100 and every result that is not 100:100 is significant statistically I'm also not even sure where you get your 40% for white >and 30% for black, where are these numbers coming from. I do not know the probability for white and the probability for black and I must do an assumption about it if I want to check if the result is significant. I found that white got something close to 40% in the ssdf games and black got something close to 30% >Still either way I never claimed the results significant just posted them out of >interest. I agree that the results are interesting The result are not signficant statistically but suggest the conjecture that chessmaster6555 is stronger than chessmaster6000 If the result after 400 games on the same computer will be 220:180 it will be enough to accept this conjecture Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.