Author: Uri Blass
Date: 14:04:24 01/03/02
Go up one level in this thread
On January 03, 2002 at 16:16:51, Robert Hyatt wrote: >On January 03, 2002 at 11:14:33, Uri Blass wrote: > >>This is analysis of Crafty under chessbase GUI >> >>New position >>[D]2r1k1r1/2p2pP1/1p5p/p3nP2/3N3B/P1P5/2q3PP/R3R1K1 b - - 0 1 >> >>Analysis by Crafty 18.12: >> >>4...Qxc3 5.Rxe5+ >> -+ (-2.05) Depth: 1/6 00:00:00 >>4...Qxc3 5.Rxe5+ Kd7 6.Re7+ Kd6 >> ยต (-1.16) Depth: 2/6 00:00:00 >> >>I am interested to know what is the remaining depth of Crafty after >>every move at depth 2. > > >I am assuming that when you start the above analysis, second line of output, >the depth is "2". > >If so, the initial depth is started at 2.5 (I kick in 1/2 ply so that the first >few fractional extensions will work). > >after Qxc3, the depth=1.5 > >after Rxe5, the depth is still 1.5 as I extend when checking rather than when >getting out of check to solve a problem of dropping into the q-search in check >which is bad if you don't do checks there. > >If the move Kd7 is the only legal move, then the depth after it will be 1.25 >as it is extended 3/4 ply and 1.5 - 1 +.75 leaves the depth at 1.25 > >After The next check, the depth is still going to be 1.25 > >After getting out of check, the next depth will be .25 which is < 1.0 and >the search drops into the quiescence search and terminates there... > >Did that help??? Yes Thanks I wanted to understand the reason that Crafty solved WAC265 one ply before my program that for now has no pruning rules and only check extensions and came to the consequence that Crafty can see at depth 2 more than my program in this position some plies after the root position. It supports also my opinion that it is possible to improve tscp by simply changing the place of lines without adding code (an idea that I suggested to tom kerrigen) Note that tscp main line at depth 2 should be only 4...Qxc3 Rxe5+ inspite of the fact that tscp has check extension(it's only extension) The reason is that TSCP does not extend checks when it is in depth 0 so after Qxc3 Rxe5 the search of tscp drops into qsearch. I think that it is better for tscp to extend checks also when the remaining depth is 0. The only problem with it is that tscp may be slower in nodes per seconds because calculating if the king is in check is not done by it in an efficient way but inspite of it I think that it can help often to see imprtant lines 1 ply deeper and it is more important than the demage of being probably 20% slower in nodes per second. 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.