Jump to content

  • Log In with Google      Sign In   
  • Create Account

#ActualÁlvaro

Posted 04 July 2013 - 07:14 AM

You should probably write a separate search function for the root, where you can implement iterative deepening, time control, returning a move, informing the user of the progress of the search...

 

I have nothing against keeping two depth variables (although my minimax searchers don't typically need both). What I was saying is that the depth stored in the hash table should be the depth left from here to the end of the tree, not the distance from the root.

 

EDIT: Oh, and you should definitely fix odd behaviors (like not winning right away when it is possible) before you add transposition tables, because transposition tables are hard to get right and make things harder to debug. You may need to make the winning score be something like (LARGE_NUMBER - length_of_the_game).


#1Álvaro

Posted 04 July 2013 - 07:12 AM

You should probably write a separate search function for the root, where you can implement iterative deepening, time control, returning a move, informing the user of the progress of the search...

 

I have nothing against keeping two depth variables (although my minimax searchers don't typically need both). What I was saying is that the depth stored in the hash table should be the depth left from here to the end of the tree, not the distance from the root.


PARTNERS