Author: leonid
Date: 18:19:53 10/06/99
Go up one level in this thread
On October 06, 1999 at 20:16:29, Dann Corbit wrote: >On October 06, 1999 at 18:47:43, leonid wrote: >[snip] >>Partially your response is just for me. When your are trying to see if the >>program found correctly all the the possibilities for best move in given >>position, it is already what I am looking for. But this work 100% only for the >>mate positions. For other positions it is problematic. >Mate positions are probably the worst. The only way to get a sure "best mate" >is to use a tool like Heiner Marxen's checkmate sovlver "Chest" which is not >available quite yet. (Others are but they do not work as well). Many chess >programs will find a mate but not the closest ones. And "usually" a mate solver >will solve much more quickly than a chess playing program. The exception is >tablebase files, which might report "Mate in 127" instantly. My question is not if my logic make mistake in the mate search but only its speed. I only wanted to say before that when the position have mate in it we can say for sure if the logic, that search it, do its work or not. If mate is in 5 moves but was found in 4, logic is buggy. If one logic do find the mate in 10 seconds and the next one in 30sec, the first is the speediest one. Case with the mate is the simplest one. Problem is with the logic that do work beyond the mate finding. How there you can compare the speed of two logics? They simply can give different results because in two games diferent value was attached to the same piece, and so on. Big mystery is for me the second part. There I still don't know where I am. I am still trying to find it on my own, or find the response on somebody else page. >>Will not say any more on >>the subject (if not I can speak for too long). I am looking for data that can >>say me more about the speed of some games, in order of recognizing mine. This I >>could do if in your data you have something like this: >> >>1) Speed at which the mate was found when that time is longer that 20 seconds. >>Search should be done by "brute force". This can revel the speed of the game, >>or at least the speed of the part of chess game that solve the mate. >The checkmate finding problem suites (such as BWTC) are among the worst as far >as bugs go. Examples -- some problems will say "the best move is <foo> and is >mate in 8", but really you will get checkmated yourself in 5 no matter what you >do [surprisingly there are a *lot* of those]. Often, there are shorter mates or >alternative mates that the test suite was unaware of. Can we say that a move >that results directly and securely in checkmate is not the best move? > In my logic I not care about this since it is or perfect, or close to this. Mistake can only come there from the fact that it was not touched for some years. But in any moment I can go back to the version that do the work. Curiousity for me could be only the speed of other games to compare when looking for mate. >>2) Speed of the chess logic when mate is not there. This speed says about >>"positional logic" agility. Search should go at fixed depth by "brute force". >>Time for solution should be beyond 20 seconds to be easely comparable. >Most test positions are not checkmate, but rather bm (best move) which does not >lead directly to mate. > >>If you have something of this kind, please make me know. >I have something of this kind. It is in the directory: >ftp://38.168.214.175/pub/Public_CAP_Results/Apocalypse/ > >Have fun. HTH. This is just after visiting your place that I deposed this message. Why you don't have your E-mail there to contact you when somebody have a question? What I found there is that many position were indicated as verified for mate in 6 moves but had the mate as such only in 2 moves. This is too short time for me (if I understood correctly what was said there) to know the speed of the game. And did you your trying at fixed depth and by brute force search? Leonid.
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.