Author: leonid
Date: 11:31:17 10/28/01
Go up one level in this thread
On October 28, 2001 at 13:46:43, Paul wrote: >On October 28, 2001 at 12:26:12, leonid wrote: > >>>Actually, my selective (and I expect that this is the same for every other >>>program) search only promissing moves. Promissing moves are those moves that >>>give quick response in most of cases. Not necesserily the shortest response. >>>Checking moves are the most promissing. Moves that affect vital king's lines are >>>the next, and so like. Between many variations of "promissing moves" I left only >>>checking because they provide high simplicity of code and no need for big >>>look-up tables. After having the chance to write my 64 bits program (if it will >>>ever happened) I, probably, will look once again into my code to see if old >>>ideas still have sense. With each big speeding old trick for "speed against >>>depth" become useless. >> >>It came to my mind to illustrate how effective is sometime selective, even if it >>goes much longer way that brute force. In this position my selective find mate >>only at 12 moves. It is mate in 9. >> >>Actually, it is nothing more but veriation of previous mate. >> >>[D]bBQQ1nrq/b2NNq1k/K6n/PQqq1qPB/3Q1Q2/3r1qR1/3q2q1/3RQQq1 w - - >> >>Selective do the work in 12 moves, in 2 secondes. Brute force in 9, in 1 min 30 >>sec. Celeron 600. >> >>Cheers, >>Leonid. > >Yes, Pretz using my "old Leonid settings" finds a mate in 9 in 19 seconds here >(on a P3/933): > >00:19 WM9 09 Nxf8+ Qxf8 Qxh8+ Kxh8 Qee5+ Kh7 Qcxf5+ Qxf5 Qfxf5+ Qxf5 Qfxf5+ Nxf5 >Qxf5+ Kh8 Qf6+ Kh7 Qh6# > >I'm just now experimenting with some new extensions, and it's a bit quicker here >with one of those, 12 seconds. But it's very tricky to find the right >combination of extensions that work just right in all positions. I, probably, never used extentions in my mate solvings. I say "probably" since I do not remember any more exactly everything that I tried during some three years of work with my mate solver. When I wrote it, I felt me almost sure that solver is "usual" ingredient of every chess program. Searching for mate, in my mind, was looking for bugs in move generator and in speeding it. Feel me happy that I did this way. When I will be ready for second part, first is ready. >I still do not ignore any moves, apart from what alpha-beta skips. I will try >later to be more selective about chosing moves, I think pruning can save more >time. Thanks for your positions Leonid, I'm having a lot of fun again these last >few days programming Pretz ... Pleased to see that you enjoy these positions. I don't know if you have now checking more as the first one; if not so, try. This will give you almost for sure higher speed (but you have it already there. Maybe you put your checking first) in mate and everything else. Graduation in selective search you can obtain by leaving one, two, three... plys searched by brute force before going to selective. This way make all procedure (at elast in mine) very simple. I am now came to my programming also. Wrote today very little but enough in order of seeing if what I do have some sense. It look like that for "my positions" it will make no difference but for other (more human like) could be helpful. Only tomorrow will continue my writing since now everybody is at home and this distract me. See you, Leonid. >Groetjes, >Paul
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.