Author: Uri Blass
Date: 20:21:56 12/22/00
Go up one level in this thread
On December 22, 2000 at 18:49:35, Thorsten Czub wrote: >is sometimes different to the thesis you have in mind. >but - lets see here... > >[d]2r1k2r/5pp1/4p3/ppqpP3/4bQPP/1B6/PPP2R1R/2K5 b k - > >>a4 ! is very strong (1...a4 2.Qf7 Kd8 3. Qxg7 Qe3+ -+) but Tiger (2 versions) >>doesn't see this move > >>How do perform other engines here ? > >ok - >i don't see why this is a selectivity problem. > > >first new paradigm: > >Gandalf likes a4, from the first seconds, in the moment >giving +2.68 in search 11. >1...a4 2.Bc4 O-O 3.Bf1 a3 4.Qd2 axb2+ 5.Kxb2 Bxc2 6.Ka1 > >CSTal 2.03 prefers Rc7 too and needs to come into search >10 to see that 1...a4 is a "better" move. > >d10, 131s, c8c7 c2c3 a5a4 b3c2 e4c2 c1c2 a4a3 f4f3 a3b2 >d10, 318s, >+1.96 a5a4 f4f7 >d10, 391s, +2.23 a5a4 f4f7 e8d8 f7g7 c5e3 c1d1 e4f3 > >now normal intelligent programs: > >hiarcs7.32 says -1.12 against white for 1...a4 and wants >to play a4 very early. I have seen it preferring a4 after 15", >i am not sure if earlier, but unimportant IMO. >after 2'40" it increases this score to -1.37 in search 10/29 >against white with 1...a4 Bxd5 main line. > >diep: >finds a4 too after a while > >stupid programs: > >junior6a likes a4 too. no problem for it. >-1.22 against white after 26 seconds. > >fritz6a wants to play Rc7 and arround 15" changes to a4. >later score says -1.47. > > >so we learn: the key move is not found via score, it is mainly >found by search. If it is the case then Crafty could have no problem to avoid castling but it has problems to do it. It likes a4 but changes its mind later to 0-0. >ok - at least gandalf finds it too early comparing its NPS-rate, > >the faster programs find the moves earlier. the slower ones >slower. > > >selectivity. hm. gambit-tiger gives a very high score for Rc7. >arround 1.8 pawns. i only waited until search 13... >3rd move in the main-line was : a4. > >i am looking forward to christophes answer. > >is tiger overseeing a4 due to a search-technique ?! > >:-)) It is easy to check it. Tell it to search only a4 and compare the scores. I did it for Tiger13 and found that the score when it searched only a4 was bigger. Tiger13 could see a4 with +1.70 score at depth 13 when it searched only a4 but needed depth 15 and almost 2 hours on my PIII450(192 Mbytes hash) to see a4 in regular search inspite of the fact that the score was always lower than 1.70 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.