Author: martin fierz
Date: 01:35:54 12/13/03
Go up one level in this thread
On December 12, 2003 at 20:43:25, Dann Corbit wrote: >In solving WAC at 5 seconds per position on a 950 MHz AMD Athlon machine, Crafty >solves 294 out of 300. That is pretty run of the mill as far as top engines go. > >What is really interesting is how quickly the remaining unsolved 6 problems are >solved. I would be amazed if any engine can match it on like hardware. i would be *very* amazed if no other engine would beat it on like hardware! here's what my pathetic engine (muse 0.876) does on a pathetic computer (P4 1.4GHz, probably similar to what you were using): >============================================================================== >WAC.230 solved in 17.74 seconds. Absolutlely stunning! >============================================================================== => no chance for muse to solve this. >PFGA: EPD record: 2 ID: WAC.131 > 10 6.74 ++ 1. Re8!! >============================================================================== >WAC.131 solved in 6.74 seconds. Almost solved in the 5 seconds alloted... >============================================================================== FEN: 2rq1bk1/p4p1p/1p4p1/3b4/3B1Q2/8/P4PpP/3RR1K1 w - - 0 0 depth eval time nodes kN/s QS BR NR pv 1/4 -238 0.00s 58 74.0 22 50 0 f2f3 . 2/10 -187 0.11s 557 8.4 40 63 68 f4e5 f7f6 e5f6 . 3/10 -187 0.11s 982 13.2 33 67 84 f4e5 f7f6 e5f6 . 4/17 86 0.55s 6068 15.9 31 80 84 e1e8 c8c1 e8f8 d8f8 d1c1 . 5/18 200 1.38s 37437 35.1 23 86 87 e1e8 c8a8 e8d8 a8d8 f2f3 . => WAC.131 solved in 0.55 seconds, much faster than crafty >PFGA: EPD record: 3 ID: WAC.141 > 9 13.11 ++ 1. Qxf4!! >============================================================================== >WAC.141 solved in 13.11 seconds. This one causes many engines problems. >============================================================================== > > 9 55.23 Mat06 1. Qxf4 Bxf4 2. Rxh5 gxh5 3. Rxh5 Bh6 unfortunately i don't print out when i fail high in the log file - happens after a similar amount of time as crafty. FEN: 4r1k1/p1qr1p2/2pb1Bp1/1p5p/3P1n1R/1B3P2/PP3PK1/2Q4R w - - 0 0 depth eval time nodes kN/s QS BR NR pv 1/4 -170 0.00s 9 13.0 31 0 0 g2g1 . 2/8 -280 0.00s 90 151.0 40 87 0 g2f1 e8e2 . 3/10 -195 0.05s 640 18.6 33 83 28 g2f1 g8h7 f1g1 . 4/15 -229 0.22s 3818 25.9 33 87 64 g2f1 f4d5 b3d5 c6d5 . 5/18 -217 1.15s 26762 34.7 33 88 77 g2f1 a7a5 f6e5 d6e5 d4e5 . 6/28 31989 49.32s 2325195 56.8 17 89 95 c1f4 d6f4 h4h5 g6h5 h1h5 f4h6 h5h6 . => mate score found in about the same time as crafty >PFGA: EPD record: 4 ID: WAC.163 >============================================================================== >WAC.163 solved in 5.12 seconds. A hair's breadth from regular 5 second solution >============================================================================== => no chance for muse >PFGA: EPD record: 5 ID: WAC.002 > 12 12.25 -2.03 1. ... Rxb2 2. Rxb2 c3 3. Rb6+ Kf7 >============================================================================== >WAC.002 solved in 12.25 seconds. Lots of engines struggle with this one. >============================================================================== FEN: 8/7p/5k2/5p2/p1p2P2/Pr1pPK2/1P1R3P/8 b - - 0 0 depth eval time nodes kN/s QS BR NR pv 1/2 84 0.05s 19 0.4 14 0 0 b3b8 . 2/6 89 0.05s 80 2.4 35 66 71 b3b8 e3e4 . 3/6 79 0.05s 309 7.4 18 69 69 b3b8 e3e4 h7h6 . 4/10 77 0.11s 1404 16.2 22 88 76 b3b8 e3e4 f6e6 f3e3 . 5/14 72 0.16s 4809 35.4 16 89 79 b3b8 e3e4 f6e6 e4e5 h7h6 f3e3 . 6/14 129 0.27s 11402 48.3 13 92 85 c4c3 b2c3 b3c3 e3e4 c3a3 e4e5 f6e6 . 7/20 68 1.26s 50011 45.6 13 91 84 b3b8 e3e4 f6e6 d2g2 e6f6 e4e5 f6e6 . 8/22 292 3.68s 144219 44.6 12 91 87 b3b2 d2b2 c4c3 b2b6 f6g7 f3f2 c3c2 b6c6 d3d2 c6c2 d2d1q . 9/22 282 5.38s 226837 47.1 11 92 88 b3b2 d2b2 c4c3 b2b6 f6f7 b6b7 f7g8 f3f2 c3c2 b7c7 d3d2 c7c2 d2d1q . => solved in 3.68 seconds, significantly faster than crafty >PFGA: EPD record: 6 ID: WAC.092 > 8 6.13 ++ 1. ... Bxg4!! >============================================================================== >WAC.092 solved in 6.13 seconds. Nearly solved in the standard 5 seconds... >============================================================================== > > 8 7.80 -0.59 1. ... Bxg4 2. e5 Bxf3 3. Qxf3 Nh5 > 4. Nd5 Qd8 5. Qg2 e6 FEN: r4rk1/1p2ppbp/p2pbnp1/q7/3BPPP1/2N2B2/PPP4P/R2Q1RK1 b - - 0 0 depth eval time nodes kN/s QS BR NR pv 1/6 74 0.11s 62 1.1 47 80 0 e6c4 . 2/12 78 0.17s 468 9.3 71 97 66 e6c4 f1e1 . 3/15 61 0.22s 2117 18.3 48 95 60 e6c4 f3e2 a8c8 . 4/17 61 0.44s 7073 27.2 41 96 85 e6c4 f3e2 a8c8 e2c4 . 5/24 54 4.18s 101755 47.2 48 88 77 e6c4 g4g5 f6h5 d4g7 h5g7 . 6/25 60 8.74s 234281 49.5 46 92 84 e6c4 f1e1 e7e5 d4f2 e5f4 d1d6 . 7/26 156 15.49s 475203 50.9 40 91 85 e6g4 f3g4 f6g4 c3d5 g7d4 d1d4 f8e8 d4b4 a5b4 d5b4 . 8/26 122 22.91s 713057 49.5 37 92 87 e6g4 f3g4 f6g4 c3d5 g7d4 d1d4 f8e8 h2h3 g4h6 d5b6 a8d8 . => solved in 15.5 seconds, again, the fail high was a bit earlier. note that 230 and 002 are very much about evaluation of advanced connected passers, not really about tactics. in games, crafty kills my engine all the time, it scores around 90% or so. i'm also impressed by crafty - but not because of what it does in these test positions :-) >I'm certainly impressed. Time to try some tougher sets and see what happens. try ECMGCP and compare it with ruffian? cheers martin
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.