Author: James T. Walker
Date: 18:55:47 11/11/05
Go up one level in this thread
On November 10, 2005 at 10:07:31, Robert Hyatt wrote: >On November 09, 2005 at 14:58:51, James T. Walker wrote: > >>On November 09, 2005 at 13:51:57, Robert Hyatt wrote: >> >>>On November 08, 2005 at 08:16:14, James T. Walker wrote: >>> >>>>My Crafty 20.1 cannot find the move either. Maybe we got our copy from the same >>>>place? It's a shame I cannot get a decent copy of the Crafty ".exe" files >>>>anymore. The one test I ran on this version plays very bad. It seems to be a >>>>giant step down in strength. >>>> >>>>00:00:14.2 0.01 12 33446835 h5 Qe7 g5+ Kxh5 Rh2+ Kg4 Rg2+ Kh5 Rh2+ >>>>00:00:47.2 3.09 12 114438836 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Kh5 Qf3 Kh6 h5 Kg7 Rf2 f5 >>>>exf5 >>>>00:00:47.2 3.09 12 114438836 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Kh5 Qf3 Kh6 h5 Kg7 Rf2 f5 >>>>exf5 >>>>00:01:08.6 4.14 13 165426487 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Kh5 Qxf7 R8e7 Qf3 Kh6 h5 >>>>Kg7 hxg6 hxg6 >>>>00:01:12.9 4.14 13 176097146 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Kh5 Qxf7 R8e7 Qf3 Kh6 h5 >>>>Kg7 hxg6 hxg6 >>>>00:02:01.9 4.59 14 292810592 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Ne3 h5 Nxg4 hxg6+ Nxh2 >>>>gxf7+ Kh5 fxe8Q+ Rxe8 Kxh2 Re5 Qf7+ Kh6 Qxc7 Rxe4 Qxa7 >>>>00:02:15.0 4.59 14 325457944 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Ne3 h5 Nxg4 hxg6+ Nxh2 >>>>gxf7+ Kh5 fxe8Q+ Rxe8 Kxh2 Re5 Qf7+ Kh6 Qxc7 Rxe4 Qxa7 >>>>00:03:14.3 4.82 15 467449473 Rg5 Qxg4+ Rxg4 Rge8 Rh2 Ne3 h5 Nxg4 hxg6+ Nxh2 >>>>gxf7+ Kh5 fxe8Q+ Rxe8 Qf7+ Kh6 Qxe8 Nf3+ Kg2 Ng5 e5 dxe5 Qc6+ Kg7 Qxc7+ >>> >>> >>>This version has been, from the beginning, a "new direction". The eval has been >>>highly modified to make it more "logical" in its organization. The asymmetric >>>stuff has been removed. All in preparation for what a group of us are currently >>>doing, namely rewriting the code piece by piece. >> >>Thanks for the explanation Bob. I Will look forward to the "New" Crafty. >>Jim > > >Note also that this position falls into the "pathological" case. It is >sensitive to hash table size and search extensions. Modifying anything will >shift the depth at which the solution is found. The reason this takes longer >than expected probably has to do with over-extending in some cases, and finding >repetitions that get stuck in the hash table. Reducing the extensions solves it >here. Or changing the hash table size so that the key entry gets overwritten >also solves it. Etc. > >That's why looking at just one position and drawing conclusions can result in a >wrong opinion and a bad change to the code. Your explanation sounds good but I played a match between 20.1 and Deep Sjeng 1.6 and Crafty came out about 190 Elo below Deep Sjeng. That put this version so low in my database (Even below 18.3)I didn't bother testing it further. Jim
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.