Author: Will Singleton
Date: 15:14:30 04/28/99
Go up one level in this thread
On April 28, 1999 at 16:55:53, Andrew Williams wrote: >On April 27, 1999 at 15:35:06, Will Singleton wrote: > >>On April 27, 1999 at 12:28:41, Peter Kappler wrote: >> >>>r3r2k/p1bb1pp1/qp5p/3pP3/PBp2P1Q/4P1RP/2B3P1/R5K1 w - - bm Rxg7; >>> >>> >>>P.S. Chess Genius 6 on my K6-300 finds Rxg7 instantly, and after a short time >>>announces mate in 8. >> >> >>Good position! We seem to have the usual discrepancy in number of moves to >>mate: >> >>Chess Genius - Mate in 8 >>Crafty - Mate in 9 >>Hiarcs 6 - Mate in 10 (mate found in 10 seconds, likes Rxg7 right away) >>MacChess - Mate in 8 (mate found in 12 minutes, likes Rxg7 after 2 min) >> >>Amateur doesn't find the move after 2 minutes and 9 ply, liking Be7 with +3. I >>better get busy and try out the single-reply extension. Problem is, I use >>psuedo-legal moves, so I'm not sure how to tell when I only have one reply! >> >>Will > > >Hi Will, > >My program doesn't find this either. It likes Be7, but takes ages to see this. >The reason I'm responding is that I've just implemented a single-response >extension. I also generate pseudo-legal moves so I'm trying this: If I'm in >check I extend one ply. I also check to see how many of the King's eight >possible moves would be legal (I have a reasonably fast attacks_square() >function for this). If this number <= 1, I then generate all the moves, make >them and then see if I'm still in check. If only one of them gets me out of >check, I extend another ply. Oddly enough, I seem to be able to do this pretty >much "for free" (ie if I do all this and then don't extend it doesn't cost me >much in the way of NPS). I've not tested this very thoroughly yet, but it seems >to be OK. The nice thing is that I could extend fractionally when I find the >King only has one safe square and then extend another fraction if it turns out >he only way out of check. I haven't tried this yet. > >Having said all this, it doesn't help me with this particular problem. > >Hope this makes sense. > >Andrew Yes it does, thanks. The fact that it doesn't help in the given position is disappointing, but I'll let you know my results. And, re nps, my prog used to be pretty fast, but every time I add something (or take it away, like lazy-eval), it gets slower. Of course it gets better too, so maybe the solution is to get real slow, like the 500 nps Centaur. :) Will
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.