Author: Uri Blass
Date: 04:50:25 01/09/06
Go up one level in this thread
On January 09, 2006 at 07:34:53, Majd Al-Ansari wrote: >Here are some of the highlights of Rybka’s weak play. First of all I would just >like to mention that Rybka’s time management is totally absent. Most of the >games Rybka will play almost instantaneously (although amazingly strong). The >only time it seems to stop and think is when it is already lost. On most games >before the critical error is made, Rybka has almost 3 X as much time as the >other engine. Although it sounds easy to fix, it probably is not. Still all >the other engines are much better than Rybka in time management. Another thing >is that Rybka does not know under promotion and can be surprised by this. And >of course it does not know basic elementary endings such as Rook vs. Bishop or >even how drawishy opposite bishop endings are. It is absolutely remarkable that >even with all these seemingly easy to fix defects, Rybka is still able to come >out on top with a huge margin. Here are some examples. > >Tourney 5_3 128mb hash (418 games) >Hardware PIV 3.2Ghz 2GB RAM >GUI Fritz 9.ctg / Ponder OFF > > >1.Rybka 1.01 Preview 2 32-bit 44.0 - 26.0 38.0 - 32.0 38.0 - 31.0 120.0/209 >2.Fritz 9 26.0 - 44.0 39.0 - 30.0 36.0 - 34.0 101.0/209 >3.Shredder 9.1 UCI 32.0 - 38.0 30.0 - 39.0 36.5 - 33.5 98.5/209 >4.Fruit 2.2.1 31.0 - 38.0 34.0 - 36.0 33.5 - 36.5 98.5/209 > > >(16) Rybka 1.01 Preview 2 32-bit - Fruit 2.2.1 [B82] >tourney 1 Home (37.3), 29.12.2005 >[0.24;0.38] >GenuineIntel 3198 MHz W=11.8 ply; 166kN/s; Fritz9.ctg B=16.6 ply; 849kN/s; >Fritz9.ctg 1.e4 0 1...c5 0 2.Nf3 0 2...d6 0 3.d4 0 3...cxd4 0 4.Nxd4 0 4...Nf6 0 >5.Nc3 0 5...a6 0 6.f4 0 6...e6 0 7.Qf3 0 7...Qb6 0 8.Nb3 0 8...Qc7 0 9.g4 0 >9...b5 0 10.Bd3 0 10...Bb7 0 11.g5 0 11...Nfd7 0 12.Be3 0 12...Nc5 0 13.a3 0 >13...Nbd7 –0.01/1 0 14.Nxc5 0.24/12 12 14...dxc5 0.38/12 17 15.0–0–0 0.09/11 13 >15...b4 (Be7) 0.12/13 16 16.axb4 0.12/11 4 16...cxb4 0.10/13 15 17.Ne2 0.08/11 7 >17...Nc5 (Bd6) 0.08/13 18 18.Kb1 –0.04/11 14 18...Rc8 (Be7) –0.04/12 13 19.Bd4 >(Nd4) 0.06/11 9 19...Qc6 –0.06/12 8 20.Be5 (Bxc5) –0.05/11 7 20...Nxe4 –0.35/12 >12 21.Nd4 (Qe3) 0.35/11 9 21...Qd5 –0.19/11 9 22.Rhe1 (Qe2) 0.50/11 11 22...Nc3+ >(Nc5) 0.68/13 35 23.bxc3 1.15/12 2 23...Qxf3 0.79/14 10 24.Nxf3 1.15/12 2 >24...Bxf3 0.85/15 13 25.Bxa6 1.20/13 3 25...Bxd1 0.98/15 11 26.Bxc8 (Rxd1) >1.08/14 15 26...Bf3 0.84/15 8 27.cxb4 (Re3) 1.16/14 9 27...Bxb4 0.87/14 8 28.Re3 >1.31/14 12 28...0–0 0.94/15 13 29.Bxe6 1.31/13 3 29...Bg2 (fxe6) 0.94/14 11 >30.Bf5 (Bb3) 1.36/12 6 30...Ra8 (Rd8) 0.98/13 6 31.Be4 (Kb2) 1.45/13 6 31...Bxe4 >0.94/13 2 32.Rxe4 1.45/13 3 32...Bc5 1.07/14 6 33.Rc4 1.56/13 17 33...Bg1 (Be7) >1.10/14 8 34.h3 1.67/14 9 34...Rd8 (Rf8) 1.14/14 10 35.c3 (Rb4) 1.66/13 21 >35...Be3 (Bf2) 1.17/14 6 36.Kc2 (Rc7) 1.71/11 2 36...Kf8 (Ba7) 1.27/13 5 37.Kb3 >(Rb4) 1.71/13 6 37...Ba7 (g6) 1.21/14 10 38.Re4 (Bd4) 1.74/13 28 38...Bc5 >1.27/13 11 39.Kc4 (Bd4) 1.74/14 8 39...Bd6 1.35/14 10 40.Bd4 1.93/14 7 40...h6 >(g6) 1.40/14 7 41.gxh6 2.34/14 6 41...gxh6 1.49/14 8 42.f5 2.48/13 5 42...Ba3 >1.73/14 4 43.Kb3 (f6) 2.48/13 6 43...Be7 1.78/14 5 44.f6 2.42/14 9 44...Bd6 >1.74/16 8 45.Kc4 (Rh4) 2.52/13 8 45...Kg8 (Rb8) 1.90/15 4 46.Rg4+ 2.55/13 4 >46...Kh8 1.94/16 7 47.Rg7 2.80/15 11 47...Rd7 2.02/17 6 48.Kb5 2.65/16 15 >48...Bf4 1.95/17 5 49.Rg4 2.69/15 8 49...Be5 (Bh2) 2.17/17 11 50.Bxe5 (Kc5) >2.88/15 5 50...Rd5+ 1.92/18 5 51.Kb6 (Kc6) 3.91/17 6 51...Rxe5 2.18/17 3 52.c4 >4.35/16 7 52...Re6+ 2.35/17 5 53.Kb5 4.35/17 5 53...Rxf6 3.49/18 20 54.c5 >4.41/16 2 54...Rf5 3.56/18 6 55.Kb6 4.41/17 3 55...Rf3 (h5) 3.73/18 6 56.c6 >4.41/14 3 56...Rxh3 (Rb3+) 3.92/17 3 57.Kc5 4.47/14 6 57...Rc3+ (h5) 4.03/18 5 >58.Rc4 4.42/16 3 58...Re3 (Ra3) 4.23/19 6 59.c7 4.42/14 1 59...Re8 4.01/15 3 >60.Kd6 (Rd4) 4.42/15 3 60...Rc8 (Kg7) 4.23/15 3 61.Kd7 4.42/15 4 61...Ra8 > > >[D]r6k/2PK1p2/7p/8/2R5/8/8/8 w - - 0 62 > >Rybka has completely outplayed Fruit and is probably in a position that most >programs (or humans) would give up by now. This is a position that even a 500 >ELO player would win against Kasparov. Simply Queen the pawn with check and then >it is a simple mop up of the pawns and a simple Rook mate. I think that you probably overestimate 500 elo players. I doubt if they even know to win with king and rook against king. 500 elo players are probably beginners who play their first or second game because I know no player with that rating who play in tournaments and based on my memory even 1000 elo chessmaster personalities often give pieces by 1 ply mistakes like a beginner. 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.