Author: Adrien Regimbald
Date: 22:59:15 08/12/00
Go up one level in this thread
Hello,
>I can detect that there is no mate in 1 for white without searching because one
>piece cannot cover all the 9 squares.
>
>I do not need to search all possible moves of white to know it.
>Chest use these kind of ideas and this is the reason that it is the best mate
>solver.
Of course I know there are techniques like this to cut down on things for a mate
solver. I was not speaking of whether it was POSSIBLE or not, I was working
under the assumption that we were talking of a normal search. This is sort of
like me saying you can't juggle 10 balls at once under the assumption that
you're on the back of a 747 aircraft, and you replying with "But a certain
juggler can juggle 10 balls!". I _know_ you could find jugglers which can
juggle 10 balls, but it's completely irrelevant to the issue at hand.
There are two problems here:
- If you do tricks like with chest to eliminate lines, you might miss good lines
to play for material/positional reasons.
- If you only look for good lines to play (a la alpha-beta or MTDF or whatever
your favourite method is) you will miss lines that are possibly mating.
The claim was that he was doing a depth <x> search and guaranteeing he saw all
mates with only a 1% penalty. The problem is that most typical searching
methods leave out huge portions of the tree. I will agree that you can quickly
check for short mates along the branches which you actually search, but I must
disagree on the issue of being able to see ALL possible mates up to depth <x>
with only a 1% penalty. This would mean that if you had a tree like this:
o
/|\
o o o
/ < x >
o
/|
o o
/ \ \
a b c
A normal search would typically see a, b, c, and skip almost all of x (well, of
course this is a gross over simplification, but you know what I mean). Call
this search a 4 ply search (no q-nodes). In order to determine all mates
possible up to 4 ply for this position, you'd have to be exploring a great
portion of x, which sort of ruins the whole approach to computer chess search,
don't you think? (well, I know it would be possible to write an algorithm to
determine that some of the moves in the later 2 branches made it impossible to
mate within 4 ply, but I'm thinking that this would come at a LOT more cost than
1%)
I have yet to see a program which will do it's normal search to depth <x> and be
able to tell you with absolute certainty whether there is a mate within that
depth available. You could easily write one which performs such a task, but
once again, that is not the issue, the issue is doing that same task while also
calculating high calibre chess moves.
Regards,
Adrien.
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.