Author: Vincent Diepeveen
Date: 04:48:04 04/16/03
Go up one level in this thread
On April 15, 2003 at 19:00:07, Nathan Thom wrote: >On April 15, 2003 at 10:23:01, Vincent Diepeveen wrote: > >>On April 15, 2003 at 08:03:36, Nathan Thom wrote: >> >>>My engine is giving me headaches :) Its performing pretty badly against test >>>positions (even thought it can beat me!). I have concerns over the following >>>statistics: >>> >>>1. The PVS search success ratio seems extremely high, of 180k attempts, only 9 >>>fail (ie value > alpha && value < beta). >>> >>>2. When writing to the hash table, of 50k writes 40k are lower bounds, 10k upper >>>bounds, 20 exact. >>> >>>Does anyone else have any statistics for what I should be getting, or any >>>suggestions where to look for errors??? >> >>what position and depth did you test? > >For these particular numbers i used > >[D]4r1r1/bB4p1/8/2p1kPKn/n7/3R4/3P4/2B5 w > >and searched to a depth of around 6 or 7 i think. Took 0.13 seconds to start all 1 other processes out of 2 00:00 1469 1k 0 0 1469 (2) 2 (0,0) 0.000 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Rd7-d5 00:00 154880 155k 0 0 7744 (2) 3 (0,6) 0.000 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Rd7-d5 00:00 207070 207k 0 0 20707 (2) 4 (0,15) 0.000 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Rd7-d 5 00:00 244484 244k 0 0 46452 (2) 5 (0,32) 0.000 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Rd7-d 5 00:00 235494 235k 0 0 89488 (2) 6 (1,64) 0.000 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Rd7-d 5 00:02 164338 164k 0 0 433853 (2) 7 (14,143) 5.941 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 Bc 1-a3 c5-c4 Ba3-d6 Ke5-d4 Bd6-f8 Kd4-e5 Rd7-d5 Ke5-e4 Rd5-d8 Ke4-e5 Rd8xe8 Ke5-d4 Re8-d8 Kd4-e5 Rd8-d5 Ke5-e4 Rd5-d7 Ke4-e5 Bf8-d6 Ke5-d4 Bd6-b8 Kd4-c5 Bb8xa7 Kc 5-b4 Kg5xh5 00:03 202307 202k 0 0 793047 (2) 8 (31,333) MATE16 Rd3-d5 Ke5-e4 Rd5-d7 Ke4-e5 B c1-a3 c5-c4 Ba3-d6 Ke5-d4 Bd6-f8 Kd4-e5 Rd7-d5 Ke5-e4 Rd5-d8 Ke4-e5 Rd8xe8 Ke5-d 4 Re8-d8 Kd4-e5 Rd8-d5 Ke5-e4 Rd5-d7 Ke4-e5 Bf8-d6 Ke5-d4 Bd6-b4 Kd4-e5 d2-d4 c4 xd3 Rd7-e7 Ke5-d4 Re7-e4 Hmm not sure such positions are good ones to test with. In general mate in x positions are very bad to generate statistics as they are unreliable and depending upon different move ordering and different branching factors and more upon extensions and quiescencesearch. So testing anything at it except whether your extensions work well should be avoided.
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.