Author: Uri Blass
Date: 15:21:50 02/15/03
Go up one level in this thread
On February 15, 2003 at 17:34:49, Drexel,Michael wrote: >Uri, you could add to the readme file something like that: >Movei runs in all GUIs which support UCI and not winboard. >Malins WB2UCI adapter is recommended. >Winboard adapter from Chessbase is not recommended. >You could also give a proposal for an WB2UCI engine file. > >two remarks: >Movei like other winboard engines shows no pv if there is only one move >available. >Movei does not announce a mate. Instead he shows something like +99.96, >+99.97,+99.99 > >I played a 10 game blitz test match on Athlon 1400 Mhz against CB native Comet >B55; both with Fritz8.ctg and ponder off. >Public Movei lost 3.5-6.5, not so bad. >Afterwards I went quickly through the games. >Movei played the middle game often much stronger than Comet but spoiled >everything later on in the endgame. The endgame treatment is very weak. I know >that Public Movei has no passed pawn knowledge and TB support but there are >other reasons too. >Movei could probably gain about 200 ELO if it would be stronger in the endgame. I know that endgames is one of the main weaknesses of Movei. My guess is that it can earn 100 elo from endgames and 200 elo seems too high even to me. I know that passed pawn knowledge is only part of the problem and another part is that Movei does not use hash tables for pruning so it is outsearched. Latest Movei give bonus for passed pawns that are protected by pawns and punish weak pawns but it still have problems in the endgames. One problem is that it does not know that it need the king to support the pawns and go to centralize the king and the result is that it could miss a win against List4.61 and exchess inspite of having advantage of 2 connected pawns(a-b pawns) Against List4.61 it was KRPP vs KR and against Exchess it was KBNPP vs KR. I know about more problems of Movei in the endgame. > >Finally some positions: > >[D] b5k1/4bp2/p5pp/8/1P1NP1P1/P1NK3P/8/8 b - - 0 45 > >Movei lost this position with white pieces. It underestimated the bishop pair. > >[D] 1r5k/4q3/r7/1pbBp1pp/3nPp2/p2Q1P1P/P2R1BP1/2R3K1 b - - 0 43 > >Movei had advantage but played the horrible 43...h4 and lost. > >[D] r5k1/pb3pp1/1p1rp2p/8/P1Pq1P2/R2B4/1P4PP/1Q3R1K b - - 0 22 > >Movei evaluated this as -0.3 for black. All my engines see clear advantage for >black. Comet had no problems to win this position with black. Movei0.0799 has no king safety evaluation and latest Movei has only small king safety evaluation. I see that latest Movei0.079992 evaluates depth=10 +0.35 a8d8 b1c2 g7g6 a3c3 b7c6 c3a3 h6h5 c2e2 h5h4 b2b4 Nodes: 3097879 NPS: 175616 Time: 00:00:17.64 depth=11 +0.59 a8d8 b1c2 d4c5 d3h7 g8f8 h7d3 c5c6 f1f3 c6c7 f3f1 d6d4 f4f5 e6f5 f1f5 Nodes: 26996520 NPS: 177737 Time: 00:02:31.89 depth=12 +0.78 a8d8 b1c2 d4f6 c2e2 d6d4 e2e3 b7g2 h1g2 d4d3 a3d3 f6g6 e3g3 g6d3 g3d3 d8d3 Nodes: 69028492 NPS: 175770 Time: 00:06:32.72 Comparison 0.0799 on the same hardware depth=10 +0.32 f7f5 b1c2 a8c8 c2e2 b7c6 b2b4 c6a4 b4b5 g8h8 d3f5 d4c4 e2c4 c8c4 Nodes: 3392507 NPS: 154979 Time: 00:00:21.89 depth=10 +0.32 f7f5 b1c2 a8c8 c2e2 b7c6 b2b4 c6a4 b4b5 g8h8 d3f5 d4c4 e2c4 c8c4 Nodes: 4952267 NPS: 157015 Time: 00:00:31.54 depth=11 +0.32 f7f5 b1c2 a8c8 c2e2 b7c6 b2b4 c6a4 b4b5 g8h8 d3f5 d4c4 e2c4 c8c4 Nodes: 15658796 NPS: 155777 Time: 00:01:40.52 depth=12 +0.41 f7f5 b1c1 a8c8 a3c3 b7c6 b2b3 g7g6 d3e2 d4d2 c1d2 d6d2 f1e1 Nodes: 25268103 NPS: 153914 Time: 00:02:44.17 depth=12 +0.41 f7f5 b1c1 a8c8 a3c3 b7c6 b2b3 g7g6 d3e2 d4d2 c1d2 d6d2 f1e1 Nodes: 30630686 NPS: 155936 Time: 00:03:16.43 depth=13 +0.36 f7f5 b1c1 a8c8 a3c3 d4c5 d3e2 b7c6 b2b3 c6e4 e2f3 e4f3 c3f3 d6d8 Nodes: 56584998 NPS: 155692 Time: 00:06:03.44 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.