🚀 go-pugleaf

In article <1993Jan25.214918.18646@spock.retix.com> dspencer@spock.retix.com (David Spencer) writes:
>In article <JON.93Jan25110340@cas.org> jon@cas.org (Jon Vander Hill) writes:
>>
>>One the the more expensive routines in a chess programs is the
>>move generation function.  The transposition table limits the
>>number of times that move generation is called to once per position.
>				                   ^^^^^
>Um, I don't think that this is so.  From the aritcles I've read and
>the source code I've seen, the "typical" transposition table entry
>has the best move for a position, but *not* all the pseudo-legal
>moves for that position. So it's possible that the stored bounds
>and stored move doesn't cause a cutoff, thus the legal moves for
>the current position may have to be generated *again*.

Another thing to note is that modern programs doing deep searches will
often fill up their transposition tables.  Then you either have to leave it
the way it is or start throwing some positions away.  I remember reading
somewhere that 1 million entry tables were too small.

+++
Lloyd Lim                    Internet: lim@cs.ucdavis.edu
Lim Unlimited                America Online: LimUnltd
330 W. Iris Ave.             AppleLink: LimUnltd
Stockton, CA  95210-3738     CompuServe: 72647,660