Author: Larry Griffiths
Date: 02:48:06 02/25/99
Go up one level in this thread
On February 25, 1999 at 05:10:19, Ulrich Tuerke wrote: > >On February 25, 1999 at 00:35:29, Don Dailey wrote: > >>On February 24, 1999 at 17:53:05, Larry Griffiths wrote: >> >>> >>>I have read a few chess papers about hashing. The ones that I have read >>>recommend not hashing at the leaves of the tree, but I seem to get better >>>performance doing it this way. (At least in the opening and middlegame). >>>Am I missing something or is this recommendation from the stone ages? >>> >>>Larry ~(:-> >> >>Do what works best. I hash 4 ply into the quies search but no farther. >>Why? Because I ran a lot of timing tests and this was optimum. > >Hashing in quiescence seems a little bit tricky to me. Imagine that you are >generating captures only but you get a non-capture move from the hash table >probe. >How do you handle this case, ignore this hit or make this move ? > >Uli > >> >>Your mileage may vary! >> >>- Don Thanks Don and Uli, I actually only hash that last plys of the primary search. I still generate all moves so the hashing is cutting out a lot of those moves. I know that some folks have code that only generate capture move's or quick gen code that validates the killer move and only generate the rest of the moves if the killer does not cause a cut-off. As to the question of quiesence hashing, I do not see how this particular case of a capture would cause a problem for me since I use a bitboard in my hash table to detect collisions and the positions would not match. My mileage may vary. How true Don. I realize that no two chess programs are the same and what may work for one may not produce the same results for another. I would not want to stop someone from trying what I have already tried, just because they thought It was a dead-end. I believe in confirming the results that have been reported by others. Larry :-)
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.