Author: Robert Hyatt
Date: 08:42:47 08/19/04
Go up one level in this thread
On August 19, 2004 at 09:37:41, Albert Silver wrote: >This is really getting tiresome. I ran a series of games between Pro Deo S&D and >Ruffian 1.05 in Arena. Pro Deo is installed as a UCI engine which shouldn't be >an issue. Arena doesn't freeze or anything so that's not the problem. The >problem is that the analysis between Pro Deo running under Arena and Fritz, >using the exact same files (personality, etc) produces different results. As a >tie-breaker, I took a look at Chess Partner 5.3 and it produced different >analysis, not to mention different depths and times. To make things even harder >to udnerstand, at least for me, when I change the hash sizes, I not only get >different times to reach a certain depth (expected), I get different main lines >adn evaluations (not expected). Note in particular the analysis at ply depth >11-13. That last part is not unexpected. Changing the hash size changes the hit rate and specific hits that occur. Changing _anything_ inside these large trees can change the score, the PV, the time, or any combination of these... > >BTW, don't think it may be using the Rebel.eng profile instead, since I checked >by renaming the default profile and made the Rebel.eng profile identical to my >personal settings. Just in case. > > Albert > >Here is an example in case. BTW, I disabled all learning, and the setting is to >now clear the hashtables. > >[D]4r2k/7p/4rqp1/p7/Pp1PBn2/5QN1/1P3PPP/3R2K1 b - - 1 35 > >Pro Deo S&D under Arena (40Mb hash): > >11 00:10 -0.64 Re6d6 d4d5 Re8d8 Qf3e3 Nf4xd5 >11 00:14 -0.72 Re6e7 d4d5 Re8f8 b2b3 Nf4h3+ g2xh3 Qf6xf3 Be4xf3 Rf8xf3 d5d6 >Re7d7 Ng3e4 Rf3xb3 >12 00:32 -0.69 Re6e7 d4d5 Nf4xd5 Be4xd5 Qf6xf3 g2xf3 Re7e1+ >13 01:23 -0.57 Re6e7 d4d5 >13 02:53 -0.76 Re8f8 d4d5 Re6e8 >14 03:45 -0.55 Re8f8 d4d5 Re6e7 b2b3 Qf6b2 Kg1h1 Re7e8 >14 05:21 -0.69 Re6e7 d4d5 Qf6h4 > >Pro Deo under Arena (60Mb hash): > >11 00:14 -0.64 Re6d6 d4d5 Re8d8 Rd1c1 Qf6e5 Qf3b3 Nf4xd5 Rc1c5 >11 00:18 -0.72 Re6e7 d4d5 Re8f8 b2b3 Nf4h3+ g2xh3 Qf6xf3 Be4xf3 Rf8xf3 d5d6 >Re7d7 Ng3e4 Rf3xb3 >12 00:34 -0.64 Re6e7 d4d5 Re7d7 Rd1d2 Re8c8 Rd2d1 Qf6e5 Qf3b3 h7h6 Be4f3 >Rd7c7 >13 01:31 -0.60 Re6e7 d4d5 Re7f7 b2b3 Qf6h4 Kg1f1 Qh4xh2 Ng3f5 g6xf5 >13 02:21 -0.60 Re8f8 >13 02:36 -0.76 Re8f8 d4d5 Re6e8 >14 03:49 -0.69 Re8f8 d4d5 Re6e7 Rd1d2 > >Pro Deo under Fritz (40Mb hash): > >35...Rd6 36.d5 Red8 37.Rf1 Qe5 38.Rc1 Nxd5 39.Rc5 Qe6 40.Rxa5 Nb6 > ³ (-0.52) Depth: 11 00:00:12 12867kN >35...R6e7 36.d5 Rf7 37.Rd2 Rc8 > ³ (-0.57) Depth: 11 00:00:29 30261kN >35...R6e7 36.Kh1 > ³ (-0.65) Depth: 12 00:01:36 101580kN >35...Rd6 36.d5 Red8 37.Qb3 Qe6 38.Qe3 > µ (-0.77) Depth: 12 00:01:42 108258kN >35...Rd6 36.Qe3 Red8 37.h3 Rxd4 38.Rxd4 Rxd4 39.Bf3 Rd8 40.Ne4 > µ (-0.89) Depth: 13 00:02:38 169928kN > >Pro Deo under Fritz (60Mb hash): > >35...Rd6 36.d5 Red8 37.Rf1 Qe5 38.Rc1 Nxd5 39.Rc5 Qe6 40.Rxa5 > ³ (-0.52) Depth: 11 00:00:13 12879kN >35...R6e7 36.d5 Rf7 > ³ (-0.57) Depth: 11 00:00:31 31116kN >35...R6e7 36.Kh1 Rf7 37.h3 Qe6 38.d5 Qd7 39.Qg4 Qxa4 40.d6 h5 > ³ (-0.64) Depth: 12 00:01:56 122630kN >35...Rd6 > ³ (-0.64) Depth: 12 00:02:02 128979kN >35...Rd6 36.h3 Red8 37.Kh2 > µ (-0.92) Depth: 12 00:02:15 142686kN >35...Rd6 36.d5 Red8 37.Qb3 Qe6 > µ (-0.74) Depth: 13 00:03:06 198623kN > >Pro Deo under Chess Partner 5.3 (60MB hash): > >00:00:08.6 0.65 10 Re6-d6 d4-d5 Qf6-e5 Qf3-e3 Re8-d8 Qe3-a7 Qe5xb2 Qa7-f7 Qb2-b3 >00:00:11.4 0.64 11 Re6-d6 d4-d5 Re8-d8 Rd1-c1 Qf6-e5 Qf3-b3 Nf4xd5 Rc1-c5 >00:00:15.5 0.72 11 Re6-e7 d4-d5 Re8-f8 b2-b3 Nf4-h3+ g2xh3 Qf6xf3 Be4xf3 Rf8xf3 >d5-d6 Re7-d7 Ng3-e4 Rf3xb3 >00:00:32.2 0.64 12 Re6-e7 d4-d5 Re7-d7 Rd1-d2 Re8-c8 Rd2-d1 Qf6-e5 Qf3-b3 h7-h6 >Be4-f3 Rd7-c7 >00:01:34.2 0.60 13 Re6-e7 d4-d5 Re7-f7 b2-b3 Qf6-h4 Kg1-f1 Qh4xh2 Ng3-f5 g6xf5 >00:02:26.0 0.60 13 Re8-f8 >00:02:40.4 0.76 13 Re8-f8 d4-d5 Re6-e8 >00:03:54.2 0.69 14 Re8-f8 d4-d5 Re6-e7 Rd1-d2
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.