Author: Mike S.
Date: 13:34:25 01/08/00
Go up one level in this thread
Hi Steve, first I want to support what you say, that testing should be done with the default settings (not talking of the hash size of course). We want to compare the "normal" performance, how it would be in a game. But I'm surprised by the result you get, although version 2000v2 seemed a little slower to me sometimes (but in general very similar), than the previous version 2000a. I use a small set of 40 positions, including 14 more ore less difficult tactical one's which I trust much. Little Goliath 2000a performs *best* in this category, in front of Hiarcs 7.32, Fritz 5.16, Fritz 3 (!), CM 6000 etc. (I don't have Nimzo 7.32 or Junior 6). My tests were mostly done on PII-333 (Hash size varied from 28 MB (LG.) to 52 for F5.16 - which is due to "testing history", available memory etc). Since you use much larger hash tables, could it be that the other programs profit much more from that? Also, did you check the LG *init*-file for the hash size? I think Li.Goliath takes the size from there and ignores the setting of the Fritz GUI. Do some cross checks also with smaller hash. I really can't believe that Li.Goliath is so far behind in tactics. Regards, M.Scheidl P.S. To allow a speed comparison: LiGoliath 2000a solved the following position in *90 seconds* (PII-333, 28 MB Hash). 2000v2 took 89 seconds; hash was set at 32 MB (maybe only 28 were used). The solution is 26.Rxh5 from a Fischer-Gligorich game: [D]2r3k1/pp1b1p2/1q1pp1p1/2r3Pn/4P3/1BN2Q2/PPP5/1K1R3R w - -
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.