Author: Rafael Andrist
Date: 05:51:32 07/19/01
Go up one level in this thread
On July 19, 2001 at 06:34:53, Tony Werten wrote: >On July 18, 2001 at 06:00:32, Rafael Andrist wrote: > >>My program missed to avoid a draw in a game (see below) due to some strange >>interactions between search and hash. I investigated it and recognized the >>following problem: >> >>- at move 33 my program recognizes a 2x and a 3x rep. during the search. The 2x >>rep. is allowed and the 3x is avoided. But in the hash, the move leading to a 3x >>rep. is stored. > >How is this different from the move leading to 2x rep ? > >Tony I see, my explanation was a bit unclear. Inside the search, there's a difference in the score stored with the move. But in the hash rests only the move without the draw score. This isn't a problem during this search (at move 33), but it causes problems in the next search (at move 35). Rafael B. Andrist >>- I get PV completly from hash, so the displayed PV is incorrect (shows 3x rep.) >>but the real PV (inside the search) is correct >>- at move 35 my porgram gets the PV leading to a 3x rep. from hash. The PV has a >>length of 8 ply >> >>---> my program cannot search to 9 ply because it has not enough time (fixed >>time per move), so it takes the move recommended from hash and draws. >>
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.