Author: Robert Hyatt
Date: 20:26:29 05/29/05
Go up one level in this thread
On May 29, 2005 at 11:17:27, Vasik Rajlich wrote: >On May 29, 2005 at 08:45:21, Uri Blass wrote: > >>On May 29, 2005 at 06:57:36, Vasik Rajlich wrote: >> >>>On May 28, 2005 at 12:41:24, Vasik Rajlich wrote: >>> >>>>On May 28, 2005 at 11:18:05, Matthew Hull wrote: >>>> >>>>>On May 28, 2005 at 10:46:48, Vasik Rajlich wrote: >>>>> >>>>>>On May 27, 2005 at 18:02:25, Uri Blass wrote: >>>>>> >>>>>>>On May 27, 2005 at 17:54:57, Pablo Ignacio Restrepo wrote: >>>>>>> >>>>>>>> >>>>>>>>Hello Thomas >>>>>>>>I would like to see playing into playchess engines room, to Dr. Robert Hyatt >>>>>>>>engines. >>>>>>>>For me is a problem to play into ICC because I have not premoves into ICC, and >>>>>>>>as a guest if there are disconnection immediately I will be lost. >>>>>>>>I would like to play a challenger (just for fun and for testing engines to be >>>>>>>>better engines and of course me too) into playchess against Crafty from Dr >>>>>>>>Robert Hyatt and against Quark, if could be possible into main room better for >>>>>>>>everyone. >>>>>>>>Unique way for me to discover new ways will be playing in long time controls. >>>>>>>>This has been a dream for me Thomas. >>>>>>>>Best regard, >>>>>>>>Pablo >>>>>>> >>>>>>>If I understand correctly >>>>>>>I understand that engines lose on time automatically at move 257 in playchess >>>>>>>and that this problem is not engine problem but interface problem. >>>>>>> >>>>>>>I think that if this problem is not fixed then engines suffer from unfair >>>>>>>conditions in playchess. >>>>>>> >>>>>>>Uri >>>>>> >>>>>>It seems that maybe you don't understand the technique. >>>>>> >>>>>>A human with pre-move (and just enough lag) will always win a closed position on >>>>>>time if the game is long enough - regardless of time control (as long as there >>>>>>is no increment), and regardless of anything that happens on move 257. >>>>> >>>>> >>>>>Bob Hyatt says crafty will never lose on time this way. So far, Pablo has not >>>>>been able to win against crafty on ICC with this technique. >>>>> >>>>> >>>>>> >>>>>>Vas >>>> >>>>Let me show you how to do it :) >>>> >>>>Vas >>> >>>Ok - crafty is damn hard to provoke into closing the position and I didn't have >>>the patience for a zillion games. >>> >>>Still - the question is, how many moves will crafty play in a 3-minute game >>>before running out of time. There is of course some limit, even if it's 1000. >> >> >>If it is 1000 then the question is if it is possible to get a game of 1000 game >>against the program. >> >> A >>>pre-mover can play an infinite amount. >>> >>>Vas >> >>No >> >>After 50 moves with no capture and conversion the program claims a draw and >>there is some practical limit for the number of moves in the game. >> >>I doubt if humans can play 1000 moves against a program without getting some >>winning position in the process and without draw by the 50 move rule. >> >> > >That doesn't contradict anything I said. The point is that you can't say that >Crafty will never lose on time. You can only say that in a 3 0 game, Crafty can >play X moves, and that X is usually going to be enough. In theory, crafty can play 180,000 moves in a 3 0 game. In reality it won't, because it will use more time up front. But in that last second of time, it can play 1000 moves. Pre-move is going to get you absolutely killed there, for reasons I have already stated. We can set up a guest account test on ICC for you to try if you want. I already know the result. _nobody_ tries that type of play against Crafty any longer. Several used to. > >> >> >>A premover can be punished by some stupid sacrifice. >>I do not think that it is very hard to make a program that detect premovers and >>punish them by that way. >> >>Of course that way have risks but if the program is not deterministic about the >>time of making the stupid sacrifice the opponent may lose most of the games >>because he will be unable to guess when the program make the stupid sacrifice >>and when to stop premoving. >> > >Yes, of course. This sort of pre-move-based guessing is normal stuff in >human-human bullet games. If you're good at it, it's easily worth 100 rating >points. > >Vas > >>Uri
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.