Author: Paulo Soares
Date: 15:45:27 04/17/99
Go up one level in this thread
On April 17, 1999 at 16:16:14, blass uri wrote: > >On April 17, 1999 at 14:39:15, Paulo Soares wrote: > >>On April 17, 1999 at 07:17:37, blass uri wrote: >> >>> >>>On April 16, 1999 at 00:03:28, Paulo Soares wrote: >>> >>>>On April 15, 1999 at 18:18:50, Peter McKenzie wrote: >>>> >>>>>The Nolot test suite is a set of 11 very hard test positions, with the 11th >>>>>position being one of the easier ones. I tried it on lambchop last night, chop >>>>>managed to solve it in about 3000sec (50min) on my 133mhz Pentium with 32mb RAM. >>>>> The solution was found at ply 11. How do other programs go on this one? >>>>> >>>>>r1b3k1/p2p1nP1/2pqr1Rp/1p2p2P/2B1PnQ1/1P6/P1PP4/1K4R1 w - - >>>>>Solution: Rxh6 >>>>> >>>>> >>>>>cheers, >>>>>Peter >>>> >>>> >>>>Peter, >>>> >>>> All engines in a PII-300 HT=32Mb. Crafty16.6 runs >>>>as a Fritz engine! Crafty runs with a logo released by >>>>Massimo Gattari(see post number 48781). >>>> >>>>If you want the new engine with logo please send me e-mail. >>>> Time(sec) Depth Score >>>> >>>> Genius6 511 9 -1.87 >>>> Crafty16.6 600 13 -1.98 >>>> Fritz5.32 >1500 >>>> Hiarcs7.01 >1500 >>>> >>>>Regards, >>>> >>>>Paulo Soares >>> >>>Fritz5.16 time:22:53=1373 seconds >>>depth:15/38 >>>score of Rxh6 -0.50 >>> >>>Fritz5.16 used 64Mbytes hash tables under windows95 P200MMX >>> >>>Maybe Fritz5.16 is better than Fritz5.32 >>> >>>Uri >> >>Uri, >> >>It is a difficult question to answer for any program. On part >>of combinations, I dont believe that had great evolutions >>in the new versions, give a looked at the tests suites. >> In compxcomp, SSDF shows the following resulteds, >>comparing the new versions with the previous: >> >>Hiarcs ....... +43 >>Nimzo ....... +42 >>Fritz ....... +12 >>MChess Pro .... -24 > >There are some problems with these results: >1)the ssdf did not test Fritz5.16 and tested only Fritz5.01 It's a good criterion not to change the version. >2)It is possible that the +12 is a result of a better book and not a better >engine It is possible. >3)the statistical error is bigger than 12 OK >4)not all the games of the ssdf are public so it is possible that there is a >hidden error in the results. > >I found an error in 4 games of the match Rebel8(p90)-Junior5(p200) (Junior5 was >slowed down by a big factor) so the tester repeat this match but it is >impossible to find errors in most of the games because most of the games are not >public games. > >Regards, >Uri Uri, wait perfection of the SSDF is a mystake. How to play as many games without errors? Impossible. In any work that you makes margins of errors exists. In my opinion SSDF always was, without doubt, a respectable and competent organization. Summarizing, it can Fritz5.32 be weakker of Fritz5.01 in compxcomp? Yes, it's possible, may be or not may be that is the question. And "The best program of all the times?" Regards, Paulo Soares
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.