Author: Robert Hyatt
Date: 10:57:14 12/22/99
Go up one level in this thread
On December 22, 1999 at 13:05:39, James T. Walker wrote: >On December 21, 1999 at 23:01:26, Robert Hyatt wrote: > >>On December 21, 1999 at 19:49:51, James T. Walker wrote: >> >>>On December 21, 1999 at 17:48:49, Robert Hyatt wrote: >>> >>>>On December 21, 1999 at 15:38:45, James T. Walker wrote: >>>> >>>>>On December 21, 1999 at 14:54:47, Robert Hyatt wrote: >>>>> >>>>>>On December 21, 1999 at 14:18:05, Mike S. wrote: >>>>>> >>>>>>>This game is really horrible. I support what you say. Maybe programmers could >>>>>>>define a condition like "If no advancement (evaluation increase) is made for ten >>>>>>>moves, so accept a draw offer". The problem can also occur when a position is >>>>>>>theoretically won by the side which has a material advantage (which the computer >>>>>>>has and won't ever accept a draw therefore), but the computer can't find a >>>>>>>method to win. >>>>>>> >>>>>>>Btw., has anybody ever got a draw offer from a program? I haven't. I would like >>>>>>>to see this also. >>>>>>> >>>>>>>Regards, >>>>>>>M.Scheidl >>>>>> >>>>>> >>>>>>If you watch crafty play on ICC you will see it happen all the time. If you >>>>>>play it on your own machine it will also offer draws.... >>>>>> >>>>>>and accept them when it is reasonable to do so. >>>>> >>>>>I assume you are talking about "The Crafty" and not Crafty Clones. My >>>>>experience with Crafty is that it continues to play for a "Swindle" when the >>>>>tablebases tell it the game is a draw. Crafty has other annoying habits like >>>>>resigning at "Mate in 1" when it has been seeing mate for about 15 moves via the >>>>>tablebases. >>>>>Jim Walker >>>> >>>>What is annoying to you is just sound chess. It has been in _plenty_ of >>>>drawn krp vs kr endings where the kr side made a mistake and lost. No reason >>>>to offer a draw until the opponent draws the game. It has been in several >>>>mate-in-N positions where the opponent couldn't mate (one was knn vs kp where >>>>it had the kp). Why resign? Make your opponent win. Remember Kasparov vs Deep >>>>Blue round 2? >>> >>>Hello Bob, >>>Well first, you missed my point about the resigning in mate in 1. When my >>>computer is counting down from mate in 22 or something like that from >>>tablebases, Crafty is also aware of the impending mate since it uses the same >>>tablebases. So why go all the way to mate in 1 and then resign just before >>>mate? This seems like a childish act to me. Something a kid would do when mad >>>about losing. >>>I understand your reasons for not resigning in the cases you mention. >>>Jim Walker >> >> >>It works like this: Once a tablebase mate in N is found, and the mate drops >>every move, it will resign at mate-in-5... because it is convinced that by that >>point the opponent knows how to finish this off. But if the opponent 'slips' >>and goes from a mate in N to a mate in N+x, then the counter resets. And it is >>possible that at mate in 5 the opponent screwed up and reset the counter. I >>don't worry enough about this since most computers don't resign at all... And >>this doesn't happen often enough to count. IE one game a week maybe? > >Hello Bob, >Thanks for the explanation. I know that after the thousands of games you have >seen Crafty play on ICC you get a little hardened towards some of the antics of >players/operators. This is understandable to me and I have been on ICC for less >than 1 year. My account is "Semi-Automatic" in that I use CSTal-2 to run other >programs. I watch every game played so it is never unattended. I have >witnessed a few strange antics by players/operators myself. Some I can laugh at >and others make me want to strangle someone. In a way you are fortunate that >you don't watch all of Crafty's games as they are played. I'm sure you miss >some cute comments from opponents at times. :-) >Jim Walker A little known fact about Crafty: I log _everything_ and while I don't go back and read it often, I do on occasion. And it has resulted in a few +censors for obscene behavior. And a couple of noplays (one guy somehow rigged his machine to offer a draw roughly twice per second, non-stop during the game. Which has a definite impact on program performance since there are lots of context switches to read data from the server thru xboard. in one case, I added a fix to crafty so that when this one obnoxious player challenged, Crafty just _flooded_ him with tells, as fast as C/TCPIP could produce, telling him "Do not repeatedly offer draws again or you will be +noplayed". After a few thousand he got the point. :)
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.