Author: leonid
Date: 07:06:09 06/10/00
Go up one level in this thread
On June 10, 2000 at 08:33:27, blass uri wrote: >On June 10, 2000 at 06:28:27, leonid wrote: > >>On June 09, 2000 at 23:57:18, James Robertson wrote: >> >>>Insomniac 0.64 on my P233 with a 4MB hash table switches to Qb7 in 167 seconds. >>> >>>Depth Score Time Nodes PV >>> 1 470 0 41 fxe6 >>> 2 464 50 264 fxe6 fxe6 Qxe6 Kh2 >>> 3 454 220 1688 fxe6 h6 gxh6 gxh6 >>> 4 368 270 8706 fxe6 fxe6 Qxe6 Bxg7 Kxg7 Qd4+ Kg8 Qxd1 >>> 5 369 380 16231 fxe6 fxe6 Qxe6 Bxg7 Qe3 Qd4 Qxa3 Qxd1 Kxg7 Qxd5 >>> 6 375 600 42821 fxe6 fxe6 Qxe6 Bxg7 Kxg7 Qd4+ Kg8 Qxd1 Qe4 >>> 7 351 1590 131221 fxe6 Qe5 Qd7 h6 exf5 Qxg7+ Qxg7 hxg7 Kf7 Kf4 >>> 8 259 6920 629363 fxe6 f6 Rc8 fxg7 Nc3 Qe5 Nb5 Bb2 >>> 9 203 18340 1685714 fxe6 f6 Qb7 Qd2 Qa6 Qxd1 gxf6 gxf6 Qxa3 Kh2 Rc8 >>>10 203 25320 2293453 fxe6 f6 Qb7 Qd2 Qa6 Qxd1 gxf6 gxf6 Qxa3 Kh2 Rc8 >>>11> 126 167130 14779415 Qb7 Qc1 fxe6 f6 Qf7 Ne5 Qxh5 f7+ Kf8 fxe8+ Kxe8 >>>11 126 186580 16418091 Qb7 Qc1 fxe6 f6 Qf7 Ne5 Qxh5 f7+ Kf8 fxe8+ Kxe8 >>>Extensions: 910044 check, 429 mate, 154399 recapture. >>> >>>James >> >>Hi! >> >>The most interesting for me is the last line. I hardly ever look into details >>that you indicated before. >> >>Please, say me how far your extensions look ahead, or just, in general, how far >>ahead it goes. What kind of extensions find the needed move? If you can, say me >>how long it could take with your logic, without extensions and hash table go 12 >>moves ahead in brute force search. Average nodes/second for your logic or some >>other that you can reach. >> >>For me "extensions" is the scond etape. I am still not there. Efficency of >>extensions I could see in this position, since Hiarcs found needed solution in 8 >>moves search. Only those extensions are more that heavy extras for the logic as >>well. Hiarcs search with extensions 8 (other way to say those numbers is also 4) >>in 14 minutes and 58 sec. Mine take without extensions (no hash) in 9 seconds. >>Question for me is: What is the best to do? Extensions, or just next few ply >>search? >> >>Leonid. > >It is clear that searching with no extensions and no pruning for a fixed depth >is not a good idea. > >There is no top program that does it. > >Uri Extensions could be put there later when best raw speed is reached. Extension in the same time could obscure greatly raw speed of the program. What exactly "pruning" is saying here? Proposing to the ply some best possible move in order of not searching entire ply? Anyway, I don't speak about minimax search where each and every node in the ply is searched. For me "brute froce" search is the search where best idial result is all the time found. How search is done, is an another question. Alpha-beta do the work as well as minimax. Leonid.
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.