Author: rasjid chan
Date: 00:02:05 06/18/05
Go up one level in this thread
"I see no reason to ever fail hard. I don't even see how fail-soft is more bug-prone" I think some(many?) don't use fail-soft because there is almost always too long a "todo" list with high priority. If I am not wrong, after we get through with how to implement it correctly, it "should be" bug free. Maybe the method may be summarize in just a few principles to stick to, basically :- 1) when returning from any search(), return a return_type of ex/ub/lb/rep3 2) when we store a best-score for exact, fail-low or fail-high,we apply revere_type() which only reveresd the type of ub/lb. 3) when we fail-high and the "best_type" is ex and if the move happen to be the last move left(more often in QS), we upgrade the return_type from the usual lb for fail-high to ex. This may be "about all" and maybe some other finer details that most chess programers should easily know. The problem is I have not monitored the "softness" of my fail soft. If fail-soft is really soft to a fair degree, why should anyone "fall-hard". When we store scores to HT, we may have a greater LB or lesser UB which is always better. Regards Rasjid
This page took 0.01 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.