Author: Steven Edwards
Date: 05:24:52 07/01/05
Go up one level in this thread
On July 01, 2005 at 04:09:03, Ross Boyd wrote: >On July 01, 2005 at 03:17:00, Steven Edwards wrote: >>Consider again BWTC.0031: >> >>[D] 2qrr1n1/3b1kp1/2pBpn1p/1p2PP2/p2P4/1BP5/P3Q1PP/4RRK1 w - - 0 1 >For comparison, on a P3 600Mhz TRACE 1.34b needs 18secs to prefer Qh5+ >and 21 seconds to announce Mate in 10. Symbolic's 41 seconds usage did not include the 1.5 seconds needed to initialize its ChessLisp interpreter and load the Lisp source. Then again, I suspect that at least a third of the time used during the search was for narrative generation as the HTML output was 408 KB long including plenty of debugging text. That's a lot of strings that have to be formatted and stored during the entire search. >In terms of nodes, Symbolic searched 86 nodes. >TRACE needed 3,808,093... a slight difference. ;) As Symbolic matures, the node counts and running times will undoubtedly change, possibly with both growing as the program gains additional knowledge and has to consider general themes. This is acceptable as Symbolic was never intended to be a blitz player. A human like search with an average one minute per move response is still the target. >I am filled with admiration for your single-minded effort and originality. Thank you. Yet so much of Symbolic has been suggested, or at least hinted at, long ago by others, not all of whom were in the chess programming domain. The really original part of Symbolic (so far) is the development of the ChessLisp language backed up by a reasonably fast toolkit. I am sure that a similar approach using a GoLisp would be the best on an attack on go. I just might try it myself if I get the chance. >May the force be with you. I might not be accepted among the Jedi due to having spent too much time in the past on the Dark Side coding traditional A/B searchers.
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.