Author: Uri Blass
Date: 03:19:28 11/28/03
Go up one level in this thread
On November 28, 2003 at 05:47:26, martin fierz wrote: >On November 28, 2003 at 05:36:59, Uri Blass wrote: > >>On November 28, 2003 at 04:56:45, martin fierz wrote: >> >>>On November 28, 2003 at 01:21:16, Russell Reagan wrote: >>> >>>>I had this position in a game I played tonight. Analyzing it, I found that I >>>>missed a draw. Ruffian took 9 seconds and 14 plies, while Crafty took 4 minutes >>>>and 22 plies to see that it was a draw (Athlon 2GHz). Ruffian gives a draw score >>>>after two repititions, so this might be why. I'm not sure what Crafty does. >>>> >>>>[D]8/8/8/1p6/8/8/pq3P2/3k2KQ w - - 0 76 >>> >>>i don't know why your crafty needs this long. my stupid engine sees the draw >>>score after 1 minute on P4 2.4GHz at depth 15. >> >>This is the kind of positions that movei does not go deep and after equivalent >>time I only start depth 9 and at depth 9 it cannot see draw score(at the time of >>replying it is at depth 10 after more than 7 minutes on p850 and still no 0.00 >>score). >> >>The reasons for the small depth may be that I do too much extensions and the >>fact that I do not use hash tables for pruning(checks in the qsearch may be also >>expensive in this position and I will need to check if this is not another >>reason). > >i don't know what extensions you would be doing here that i'm not. i do a full >ply on every check, which is more than many suggest. since this position is >about giving lots of checks, you can imagine how many extensions i get :-) The main extensions are more check extensions so finally with 16 mbytes hash I get a draw score at depth 11. depth=11 +0.00 g1g2 d1c2 h1h7 c2b3 h7f7 b3a4 f7a7 a4b3 a7f7 Nodes: 281062308 NPS: 149922 Time: 00:31:14.72 > >i do checks in QS too, but only on the first ply, and also only if i believe >that i have an attack against the enemy king - so my program shouldn't be doing >them here. Why not There should be attack on the first ply against the enemy king. > >this also looks like a position where a hashtable will make a HUGE difference - >few pieces, no relevant pawn moves in the perpetual check line; so i suspect >that this is the reason you're not going deeper here. > >>What is the level of your engine? >weak ;-) > > >>Do you test it in games against free winboard chess engines and if you do can >>you mention engines of similiar strength? >yes, i do of course, it's very frustrating :-) > >gothmog 0.3.1 and movei 0.799 are about similar in strength - i do nunn2 matches >with blitz time controls (1 min + 5'' increment); i don't have an opening book >(yet), so i can't play regular matches. > >cheers > martin Today I also do nunn2 match to test my engine. I have a small opening book but I am not interested in working about book. In the past I mainly tested it against itself when it always changed the first move but I decided recently to test it in the nunn2 match against other engines to have some data about it's results against all the levels for comparison with future versions. 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.