Author: Dieter Buerssner
Date: 06:35:23 02/17/01
Go up one level in this thread
On February 17, 2001 at 05:39:18, Uri Blass wrote: >These tablebases are needed to evaluate the game between Deep Fritz and Germany. > >8/8/kr6/8/8/P1PK3R/6r1/8 b - - 0 1 > >In this position Germany played Rb3. > >Eduard Nemeth claimed that Rb3 was a mistake and better was Kb5 or Ra2. >Is he correct. With the help of Yace, I proved that Kb5 yields in Mate in 32 or better, also Rbg6 yields a mate in 33 or better. After Rb3 DF played: >What is the evaluation after 65...Rb3 66.Kc4 Rxa3 67.Kb4? Here I could prove a mate in 36 or better for Ra1, but it took a much longer time to prove this, and I had to look at much more variations. For proving this, I used a feature that was actually inspired by a message of you here in CCC. In Analyze mode, Yace will remember important evaluations of allready analyzed positions and subtrees. I just made moves suggested by Yace rather fast, until a mate score was found. Then I went back move for move. At points, where Yace no switched its mind, and did consider another move for White without mate score, I did the same for the alternative. After about 3 hours, the mate score was propagted back to the position after 67...Ra1. I did not care at all, to find a short mate, I just was interested in any mate score propagating back. For this position, I used a version of Yace, that probes TBs very often, even in quiescence search. The performance in nodes/s was very slow, usually under 20000 nodes/s on my AMD K6-2 475. Without TBs I would get more than ten times the nodes/s on my computer. I did this with 15 M hash and 2 M egtb_cache. I had the relevant 5-men TBs for this positions with the exception of KRRKP and the Q-tables, but they are not really needed. Of course, it may also be a proof, that I have a horrible bug :-(. But at least for the positions after 65...Kb5 and 65...Rbg6 the mate was also found by "normal" search without stepping through different variations, but needed more than four hours in both cases. By stepping through the variations, I needed less than 1/2 hour to proof the mate after 65...Rbg6. I think, it would be futile to try the position after 67. Kb4 without stepping through the variations or a special TB. Regards, Dieter
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.