Author: Don Dailey
Date: 11:49:56 12/18/97
Go up one level in this thread
>>I had a very similar idea a few months ago. As a result I have a >>positional test of 24 problems, all from great grandmaster games and all >>commented by grandmasters. I discarded positions with solutions not >>found by any programs. The main difference with your proposal is that >>instead of 30 minutes I used only 5, more or less the average amount of >>time in games played at 40:2. >> >>The test has been posted on CCR under the name "CCR Test", with results >>on a P200MMX/64. Below I post the global results of some programs and >>the test set. Please tell me what you think of it. >> >>Rebel 8/9 17 >>Hiarcs 6 16 >>Mchess 6 16 >>Rebel 6 16 >>Mchess 7 15 >>Fritz 5 13 >>Genius 5 12 >>Shredder 1 11 >>Fritz 4 10 >>CM5K 10 >> >>Enrique >I had my program Patzer run this test with 27.5 minutes per position >over >night on a Sun Ultrastation 167 MHz: > >Pos. 1 Qa4 till 6:07 then changed to Ra1 ( does this count for 5 minutes >:-) >Pos. 2 9:05 >Pos. 7 0:06 >Pos.11 16:22 >Pos.12 0:00 >Pos.13 0:00 >Pos.14 14:14 >Pos.15 0:24 >Pos.16 10:53 >Pos.18 0:00 >Pos.19 18:55 >Pos.21 0:03 > >Position 1 shows that it is not enough to wait only 5 minutes. > >I am curious: >Which positions were "solved" by which program in which time with which >eval? I just ran Cilkchess with time till solution (but requiring at least 3 ply of search in cases where it finds the move instantly.) I am going to run for the full 5 minutes next to see how many problems if any Cilkchess loses. I'm not clear on the method the other programs used when the test was run for them. Cilkchess 2.0 with 1 processor on Alpha 466 mhz Total solved = 16 1) Qa4 8.8 2) Qe3 - 3) h3 149.0 4) Bg5 - 5) Qb6 0.0 6) f5 0.0 7) Nb1 - 8) bxc6 - 9) Na5 - ; we get this one in about 7 minutes 10) Rc6 1.3 11) R1c2 49.1 12) d4 0.0 13) Qe5 0.1 14) Qa5 0.0 15) Bxd6 10.9 16) Qc8 - 17) Ke8 42.7 18) e5 0.0 19) f5 - 20) f4 - 21) b4 1.0 22) Bb5 3.5 23) Nd1 200.3 24) h4 114.5 - Don
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.