Author: Mike S.
Date: 12:07:17 05/24/00
Go up one level in this thread
On May 24, 2000 at 14:16:07, blass uri wrote: >(...) >I agree that if the human opponent is using a long time for one move(for example >more than 10 minutes at tournament time control) it may be better to stop >analyzing the reply for the expected move and start analyzing the second best >move and after it the reply for the second best move but I do not think that >this idea is very important. > >Another possible idea is if you discover by analyzing the predicted move that >the predicted move may be a mistake(an extreme example is if the evaluation is >mate foryourself after the predicted move) there is no point in continuing to >analyze the predicted move and it is better to start to analyze a different >move. From the viewpoint of an user, the latter seems to be a good idea and especially attractive , when the program furthermore discovers during the pondering, that it's mate whatever move the opponent will make. It could then "suddenly" announce a mate out of the permanent brain! I remember that the idea of a "multiple" pondering was already used once in a board computer, the SciSys (or Saitek) Turbo 16K, and also in the 24K I think. Maybe this could be decided by the program according to the evaluation differences between the best, 2nd best and 3rd best expected moves: For example at tournament setting, if the difference is smaller than +- 0.2 pawns, it could use the first minute for the 1st move, then switch to the 2nd best etc. Of course I do not know which values and numbers of moves would be the best compromise for this. Regards, M.Scheidl
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.