Author: Drexel,Michael
Date: 11:00:36 05/02/03
Go up one level in this thread
On May 02, 2003 at 11:37:00, Jorge Pichard wrote: >On May 02, 2003 at 10:19:39, Drexel,Michael wrote: > >>On May 02, 2003 at 07:13:21, Jorge Pichard wrote: >> >>>Tomorrow I will start a new mach of 10 games against version 8.0.0.23 against >>>CT15 so we can compare them against the same opponent. >> >>This is IMO of no interest. A 10 games match is almost worth nothing. >>You need a lot more games to get an estimate. >>The SSDF is testing with 40 games per match. >>A program which is clearly stronger than the opponent should win a 40 games >>match almost with probability 100%. > > > >Again this is only so the CCC members can get an Idea NOT an exact assurance >that Program X is stronger than program Y. But in this particular case where >Fritz has its origianl version 8.008 and also version 8.0023 which Tony from >the SSDF is testing right now, we are not too sure if version 8.008 is really >stronger than version 8.0023. When Shredder 7 and Shredder 7.04 were tested >against all the three versions of Fritz, it gave us a clue with a small margin >of error that version 8.0.0.8 is the stonger against Shredder, but that doesn't >indicate that it is stronger overall against the rest of the programs. >Therefore, my little test will only give you a clue. You even cant get an idea. 8.0.0.8 and 8.0.0.23 are almost equal. > >>If the weaker program is lucky and gets the right positions out of the opening >>it can win a 10 games match against a program which is clearly stronger. >> >>Michael > >It all depends how much weaker, if the weaker program is more than 50 points >weaker, than the opening might not help the weaker program much. Just to give >you an escenario lets take List 5.04 with its opening book for instance, versus >shredder 7.04 without an opening book at all, if you match them in a 40 games >match i would not put my money on List 5.04. There is one major reason why you >can't depend on opening book too much A programm can be more than 100 points weaker and can play a certain position almost perfect and the stronger program can play this position like something which is rated <1800. You play with opening book for both sides and with this scenario a 2500 program could draw or win a 10 games match against Shredder 7.04. > >1st. The opening book could have been created by collecting openings from human >players, and of course these opening are NOT guarantee to be error free. > >2nd. It is not clear if top programs such as shredder 7 or Fritz 8 are able to >create opening moves that could be better than those taken from humans players, >specially if the opening books were taken from games of players less than the >top 100 FIDE ranking.
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.