Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Cleaned ECM aka WACII revisited

Author: José Carlos

Date: 10:24:56 09/23/01

Go up one level in this thread


On September 23, 2001 at 12:45:41, Uri Blass wrote:

>On September 23, 2001 at 12:15:35, José Carlos wrote:
>
>>On September 23, 2001 at 11:02:09, Uri Blass wrote:
>>
>>>On September 23, 2001 at 08:54:02, Gian-Carlo Pascutto wrote:
>>>
>>>>On September 23, 2001 at 05:29:51, Peter Berger wrote:
>>>>
>>>>>
>>>>>[D]2b3k1/p4ppp/7q/2Q5/8/P3r1P1/1r4BP/R3R1K1 b - - bm Bb7; id "ECM.1197";
>>>>>
>>>>>This one looks wrong.
>>>>
>>>>Nope...Bb7 is the fastest mate.
>>>
>>>What is the importance of finding the fastest mate.
>>>There are programs that stop to search when they find one mate and I do not see
>>>it as important for games if they know to find a shorter mate in the next move.
>>
>>  This is wrong, partially. Of course playing programs are not mate provers, but
>>if you stop the search in _any_ mate, you can play a check that leads to mate in
>>5. The next move, check again, which leads to mate in 5. Then, another move that
>>leads to mate in 6, and so on... You might lose on time meanwhile.
>
>I wrote
>
>"I do not see it as important for games if they know to find a shorter mate in
>the next move"

  I wrote: "partially"

>It means that the program can stop in the first mate and in the next move to
>stop only when it finds a shorter mate.

  I agree here.

>If the program can remember the full tree that proved the mate then it even does
>not need to use time in the rest of the moves to find a shorter mate in the next
>move.

  Agreed again.

>If the tree is too big to remember for the program
>then it means that the program needed a lot of time to find the mate and it
>means that it had time and again there is no problem of losing on time unless
>the program got into big time trouble in the last move and getting into a big
>time trouble in one move is usually not logical(the only exception is if the
>program thought that it is losing but it is not the case here).
>
>Uri

  I was only trying to point that it is dangerous to stop when a mate is found
withot any other consideration. Of course, if you consider other things, you may
solve it easyly.
  I know what I say because this _happened to me_ the first time I wrote Averno.
As I've said before, I wrote it from scratch, and so I did a lot of stupid
stuff, until I realized it was stupid, and figure out a smarter way to do it. I
remember a game of version 0.14 where the program played Qh3+ Kh4 / Qh4+ Kh5 /
Qh5 Kg6 etc. saying mate in 3 all the time. After that, I handled it in a less
stupid, but still stupid way: remembering the pv when there was a mate score.
Then, the opponent made a move out of the pv, and the stupidity begun again.
  When you do things like this, you feel the need to prevent others from doing
the same :)
  Of course, there're zillions of ways to handle this smartly.

  José C.



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.