View Single Post
  #2   Report Post  
Old April 18th 04, 12:52 PM
hb
 
Posts: n/a
Default transposition tables...

Tommy wrote:
I have implemented transposition tables and most of time time the program
works fine and they provide great saving. However in some rare cases the
program now sees checkmate possibilities which are not there. Do you have
any idea of what I could have done wrong? Could it be a clash which also
happens to have the same zobrist key or is that too difficult to have?
I use a 32MB TT table with 1,333,357 elements (i tried to have a prime
number just to see if that was the problem....). I also tried different
sizes.
The replacement schema so far is: always replace
The random64() function is made by many rand() ORed together, but I also
tried the more advanced function in Crafty and the problem remains....
I use TT for returning exact values, trying the best move first (move
ordering) and if alpha or beta value produce cut-offs. The part which seems
to give problems is what is supposed to be the most straight forward:
returning exact values.

Thanks in advance,

Tommy


Maybe you're pruning escape routes.