Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Move Ordering

Author: Robert Hyatt

Date: 19:32:18 12/25/02

Go up one level in this thread


On December 25, 2002 at 16:57:22, Dieter Buerssner wrote:

>On December 25, 2002 at 15:18:29, Robert Hyatt wrote:
>
>Uri has commented to the pruning argument.
>
>>[Fine 70]If the tree were ordered
>>perfectly it takes 26 plies, period...
>
>I don't agree. I actually thought rather careful about this, and I think I
>understand perfectly your reasoning. I gave some other argumentation in
>http://f11.parsimony.net/forum16635/messages/39246.htm and some other messages
>in that thread. This included an experiment, where my engine finds the solution
>of Fine 70 in 22 ply (with normal moveordering). It used 26 ply with random move


Random won't cut it.  You need _worst_ ordering.  You have to (by luck)
play the worst moves for black and good moves for white.  Then you reach
(quickly) a position where white can win the pawn within N plies (I usually
solve this at 18 plies for example, but it varies version to version since
move ordering can change effectively, when the eval changes).  Then you
store that position where you can force winning the pawn, and then search
best moves for black which pushes that position out near the tip, but now
you know you can win it and you back up the right score.

This is pure luck, but best-first will fail to find the solution if you can
somehow always search the best move first, because you will not have found
the "i can force a win from here" before you actually "get here".



>ordering. Also, it used 26 ply without hash tables. The experiment was done
>without evaluation function (only material evaluation), and it was considered
>solved, when a pawn capture was found.
>
>Your argument goes like this: You need non perfect move ordering, to pick later
>up those inferior lines through the HT. My reasoning: Even with perfect move
>ordering you will pick up more than enough of those inferior lines, because any
>2ndd ply, you need to search every move anyway, and find a refutation for it.
>
>Regards,
>Dieter

Yes, but try it with best first ordering, which you can do after searching
the thing to 22 plies and finding the win.  Just search to 22 plies again with
all the hash moves kept, but the scores cleared.  No solution.





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.