Author: Michael Yee
Date: 03:59:28 03/19/05
Go up one level in this thread
On March 19, 2005 at 06:40:50, Reinhard Scharnagl wrote: >Well, I repeat something I have already written. There is no need to tell the >GUI which ponder move is intended to answer. An unfair GUI will have the chance >to give that information to a friendly engine. A pondering machine should be >free, what it is doing with the in between time. It could use it to calculate an >answer to its internal ponder move, but e.g. also build up temporary tables and >hints. It should be up to the engine itself, to what it would use the >calculating time. Thus communicating ponder moves is to restrictive and overmore >could be misused. > >Reinhard. Hi Reinhard, I agree with you almost 100%... The engine should be free to do whatever it wants on the opponent's time (as long as the engine was given permission to ponder). Here are a couple points that I didn't include in the spec yet: (1) the suggested move would always come from the last thinking line of the engine itself--if it's available (2) the suggested move is just a suggestion (and is there merely for people who liked the UCI convention) (3) the engine isn't *required* to tell the GUI its pondermove (with "info pondermove ..."; but some users might like to see it I'm open to removing "suggestedmove", however, since it seems trivial for an engine to come up with it itself... It seemed more necessary under UCI where the "state-remembering" responsibilities of the engine were minimized. I hadn't planned for the possibility of an "unfair"/unfriendly GUI... It seems like such a GUI could also forward the thinking lines of one engine to another. I see a (slight) tradeoff between the engine being useful (to the human user) and ensuring total fairness in engine-engine matches. Michael
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.