Author: h.g.muller
Date: 05:04:16 02/11/06
Go up one level in this thread
It doesn't sound to me like anything you should worry about. If the engine sees over-the-horizon mates, there is no guarantee that they find the fastest. Since over the horizon, the search is by definition not full width, and branches with the faster mate are pruned. Most likely the mate is only found through hits to hash-table entries that happened to have a depth much larger than the needed one, so that the depth-calculation implicitly done in the mating score adds the depths (and not through search extensions). What it finds can be very dependent on the way the hashtable is filled (ore overwritten), and changes in move ordering can cause the branch that was responsible for filling the hash entry to high depth to be pruned, making the mate disappear again. (The quick path to such a position that causes it to be searched at high depth early, is usually due to bad play of one side, so it is not surprising that alpha-beta is eager to prune it.) You'r SMP version might be slightly less efficient in pruning that branch, because it can be searched in parallel with the better branch that will become responsible for pushing it out of window.
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.