Author: Uri Blass
Date: 21:03:12 07/10/04
Go up one level in this thread
On July 10, 2004 at 21:49:04, Anthony Cozzie wrote: >On July 10, 2004 at 18:24:26, Jon Dart wrote: > >>I have recently modified Arasan so that it can display 2nd and 3rd-best >>moves for each test position. Using this I have gone through my >>previously released test suite and some other tests I have collected >>recently (mostly from Arasan games), and weeded out positions that >>don't have unique good moves. I also removed some that were too easy. >>The latest version is at http://www.arasanchess.org/arasan4.epd. >>Any comments or corrections welcome. Last I ran this, Arasan was getting >>about 95 out of 115 correct at 60 seconds/move. >> >>The last test is interesting and wasn't solved at the 60 second mark: >> >>[D] 2kr3r/Rbpnq3/1p1pp3/2P3p1/3P4/2P1P1p1/Q2N1PP1/4KB2 w - - >> >>best move Rxb7. Phalanx played this in a test game (Arasan is getting a + score >>over Phalanx overall in my tests now, but Phalanx is particularly good at >>tactical shots and is still a tough opponent). >> >>--Jon > >1. Bf1-a6 g3xf2 2. Ke1-e2 f2-f1=Q 3. Nd2xf1 d6xc5 4. Ba6xb7 Kc8-b8 5. Qa2-a6 > = (-1.41) Depth: 7/26 00:00:02.22 790kN >1. Ra7xb7 g3xf2 2. Ke1-d1 Nd7-b8 3. Qa2-a8 Kc8-d7 4. Rb7xc7 Kd7xc7 5. Qa8-a7 >Kc7-c8 6. Qa7xe7 > = (4.10) Depth: 7/26 00:00:02.52 984kN >1. Ra7xb7 g3xf2 2. Ke1-d1 Nd7-b8 3. Qa2-a8 Kc8-d7 4. Bf1-b5 Nb8-c6 5. Bb5xc6 >Kd7xc6 6. Rb7xb6 Kc6-d7 7. Qa8-a4 Kd7-c8 8. Qa4-a6 Kc8-d7 9. Qa6-b5 Kd7-c8 10. >Rb6-b8 > = MAT10 Depth: 11/33 00:01:16.88 58547kN > >Zappa is not very good at getting deep PVs, so it takes some time for the mate >score :( > >anthony zappa is good here relative to movei and most programs. I believe that it probably use mate threat extensions that movei does not use today. Movei only use check extensions and partial recaptures extensions dependent on the remaining depth and pawn push to the 7th extensions in part of the cases. I prefered simply not to try more extensions because there are positions when it cannot go deep and I think that better pruning and better order of moves is more important. When I have it I may try more extensions in the future. One advantage of pruning is that they allow more extensions without explosion of the tree. I believe that even without using hash tables for pruning there is a big area for improvement. 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.