Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Problems connected with hashing

Author: Robert Hyatt

Date: 08:40:40 09/08/00

Go up one level in this thread


On September 08, 2000 at 07:41:57, Jan Pernicka wrote:

> Hi,
>summer is almost over and I return from holidays to Internet :)
>
>I have just implemented (= and debugged) hashing in my chess program and
>have to face (at least) 2 problems:
>
>   1) How to update Best Variation when using hash?
>       My solution: Do everything as normal and when transposition is detected,
>                    trace (with help from best move saved in hash table)
>                    partial best line through chain of positions in hash
>                    table.
>

That works.  You can occasionally get a bogus move, particularly when you
consider that for every move on the PV that had a fail high, the next move
could be a fail-low node and the hash has no (or a wrong) "best move"



>  2) (more "dangerous"): When detecting 3-fold repetition (in fact 2-fold)
>     in the tree in position A - I return score=DRAW_SCORE and e.g. in the
>     position (B) immediately above position A I store in hash table score
>     which is influenced by score of position A, but this score (of A) is
>     relative to the root(it is repetition): in position B - when it will
>     occur somewhere else in the tree - it could cause errors.
>     My solution: I simply ignore this fact.

I do the same thing.  Hashing fails when you deal with three-fold repetition
and the 50-move rule.  The fix is ugly and hurts more than it helps.



>
> Thanks for your answers and have a nice day.
>
>         Jan



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.