Author: Uri Blass
Date: 01:15:35 11/09/04
Go up one level in this thread
On November 08, 2004 at 21:58:55, chandler yergin wrote: > (snip) > >On November. >Without analysis Fritz 5.32 & Crafty also found 11. h5 with no problem. > >I just did it! > >>Fritz5.32 and Crafty cannot find 11.h5 in a short time >>I checked and found that Fritz5.32 cannot find it in a reasonable time. > >They did for me.. NO Previous analysis > > >>The facts are that Crafty even cannot find 11.h5 after many hours based on >>analysis that was posted here. > >Then someone did something wrong. It may be interesting to see your analysis. > > > >>Of course it may find it if you use a special personality but I talk about the >>default personality. > >I don't use personalities.. just factory settings! someone who used Crafty19.17 analyzed it for hours and it did not give the correct result. I believe more him after seeing your analysis that is obvious result of learning. > > >> >>> >>>You were expecting this? >>> >>>Ed - Rebel 12 >>> [D]rnbq1rk1/pp1n1p1p/4p1p1/3pP1N1/1b1p1P1P/2NB4/PPP3P1/R1BQK2R w KQ - 0 1 >>> >>>Analysis by Shredder 8: >>> >>>11.Qh5 gxh5 12.Nxh7 >>> -+ (-9.90) Depth: 1/4 00:00:00 >>>11.Qh5 gxh5 12.Bxh7+ Kg7 >>> -+ (-10.55) Depth: 1/4 00:00:00 >>>11.Qh5 gxh5 12.Bxh7+ Kh8 >>> -+ (-10.38) Depth: 1/5 00:00:00 >>>11.0-0 dxc3 >>> -+ (-5.13) Depth: 1/5 00:00:00 >>>11.0-0 dxc3 >>> -+ (-4.06) Depth: 1/5 00:00:00 >>>11.0-0 dxc3 >>> -+ (-4.04) Depth: 1/5 00:00:00 >>>11.Bxg6 Bxc3+ 12.bxc3 hxg6 >>> -+ (-4.00) Depth: 1/5 00:00:00 >>>11.Bxg6 fxg6 12.Qxd4 >>> -+ (-3.54) Depth: 1/5 00:00:00 >>>11.Bxg6 fxg6 12.Nxe6 Bxc3+ 13.bxc3 >>> -+ (-3.46) Depth: 1/7 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 1/7 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 1/7 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 2/2 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 3/3 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 4/4 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 5/8 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 6/14 00:00:00 >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 7/10 00:00:00 6kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 8/8 00:00:00 15kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 9/12 00:00:00 16kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 10/21 00:00:00 31kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 11/24 00:00:00 64kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 12/24 00:00:00 170kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 13/20 00:00:01 337kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 14/26 00:00:02 690kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 15/32 00:00:05 1349kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 16/30 00:00:08 2049kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 17/42 00:00:18 5155kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 18/40 00:00:28 8195kN >>>11.h5 dxc3 >>> ³ (-0.27) Depth: 19/37 00:00:46 13986kN >>>11.h5 dxc3 12.hxg6 hxg6 13.a3 cxb2+ 14.axb4 bxa1Q 15.Rh8+ Kxh8 16.Bd2 Qxd1+ >>>17.Kxd1 Nxe5 18.fxe5 Kg7 19.Bxg6 >>> ³ (-0.52) Depth: 20/52 00:02:37 51108kN >>> >>>(Yergin, MyTown 08.11.2004) >> >>Score for 11.h5 goes down so shredder may change it's mind if you give it more >>time. > >Propably will. > > >>It clearly learned and your analysis now is different from previous analysis. > >No, that's what I got the first time.. No You should understand that it is not logical that the score and main line will be constant at depthes 1-19 > > > > >>>>>>I do not have Shredder8 but here is some analysis by Shredder7.04 on A1000 and >>>>>>it does not get h5. > >Then you need Shredder 8! No You need not to post stupid analysis that does not make sense. programs even if they can find the correct move change their mind about the main line frequently. The fact that Shredder never change it's mind is practically a proof that it did not find the move instantly. > > >>>>>> >>>>>>after learning I suspect that yace can also see it very fast so it proves >>>>>>nothing >>>>>> >>>>>>Shredder8 may be able to search deeper but not that deeper. >>>>> >>>>> >>>>>Check the Analysis and times... >>>> >>>>This is exactly the point. >>>>The analysis and the time does not make sense for a program that did not learn >>>>from analysis of the game. >>> >>>Of course it does.. It doesn't just check Main Line.. it evaluates many, >>>you just have the PV showing. >> >>I am not talking about the fact that it searches many lines but about the fact >>that I am sure that shredder 8 cannot find h5 in a short time if the user analyze >>the game so shredder updates his learning files. > >Well it did and does. It's fast too. >Why do think it or any Top Program will not consider logical Candidate moves? I do not think that top programs will not consider logical candidate moves. I am a programmer and I know that my program consider every possible legal move. I think that top programs will need time to find h5 and not do it instantly and I think that they change their mind often in the first seconds. The fact that in your analysis shredder changed it's mind first at depth 20 say something. If you give it position that it never saw before it is going to change it's mind about the evaluation frequently in the first seconds unless there is a forced mate or simple repetition that it prefers. >Also Shredder has a large Opening Tree, and it references the Database. If the program is not in book that I do not see how it relevant. 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.