Author: Robert Hyatt
Date: 06:57:39 01/18/00
Go up one level in this thread
On January 17, 2000 at 14:26:12, Lex Loep wrote: >On January 17, 2000 at 12:56:48, Robert Hyatt wrote: > >>On January 17, 2000 at 10:24:05, Ernst Walet wrote: >> >>>On January 17, 2000 at 10:12:14, Robert Hyatt wrote: >>> >>>>Someone just pointed out something that I wanted to get some confirmation on >>>>about this interface. Based on what I saw, it seems that this interface is >>>>breaking at least two things in "Crafty" if it is run under chess partner: >>>> >>>>(1) When sending a move to the engine, they first seem to send an "undo" >>>>followed by a "force" followed by the move, followed by "move". This simply >>>>tears up thinking on the opponent's time. Because the undo/force commands >>>>terminate the ponder search in crafty instantly. >>>> >>>>(2) This also totally breaks my book learning, since it is done every move and >>>>crafty never realizes that it is 10 moves out of book, since the position keeps >>>>getting reset. >>>> >>>>If this is true, it is just like the original version of the ChessBase >>>>interface, and means engine vs engine or engine vs operator games are no >>>>good, at least for the engines that are _really_ winboard compatible. >>>> >>>>Can anyone confirm this if you have chesspartner 4.2? IE play a few games >>>>using Crafty as one engine (assuming it supports engine vs engine games) or >>>>else play a few moves manually vs crafty in this interface, making sure that >>>>it gets out of book. Then send me the log.nnn file. We don't need a bunch of >>>>tournament results from this interface it if is really behaving like this, as >>>>the games are meaningless... >>> >>> >>>I think I know who told you, as he told me as well. I tested it with the >>>Rebel-Tiger beta and can confirm this happening, alas. >>> >>>Ernst-J. >> >> >>If this is true, (a) it is horrible programming and (b) means no matches played >>under chesspartner are worth anything at all. That "trick" wrecks everything I >>do for sure. I assume it is not helpful to others. But in any case, in such >>a match, crafty plays without pondering, will _definitely_ clear the hash >>frequently, and can't learn at all. > >It is true, this 'trick' was necessary to get the GUI features like movelist >navigation, variations, game analysis etc. working kind of reliable. >(a) It is not horrible programming, it works the way it was intended, I just >did not realize it would break crafty. >Problem is the winboard interface was never intended as a computer interface >and has it's short comings, not to mention the variations amongst engines. >Anyway, the winboard adapter is being worked on and this problem will be >fixed soon. > >Lex It doesn't just break crafty.. How can _any_ program think on the opponent's time, then be told to take back a move, make a move, and then search again possibly use the ponder search results? The winboard protocol is dead simple. You send the program a move, the program sends you a move, you send the program a move, etc. No need to 'force', or to back up a move, or anything else... This will break _every_ winboard compatible program, because "force' has an explicit meaning in the winboard protocol, namely to "stop searching and get ready to make a move on demand."
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.