Author: Robert Hyatt
Date: 10:30:47 04/16/00
Go up one level in this thread
On April 16, 2000 at 11:42:12, Dave Gomboc wrote: >On April 16, 2000 at 10:11:13, Enrique Irazoqui wrote: > >>On April 16, 2000 at 10:05:45, Robert Hyatt wrote: >> >>>On April 16, 2000 at 04:33:18, Jason Williamson wrote: >>> >>>>On April 15, 2000 at 23:22:47, Robert Hyatt wrote: >>>> >>>>>On April 15, 2000 at 13:59:26, Tony Hedlund wrote: >>>>> >>>>>>On April 14, 2000 at 12:54:59, Stefan Meyer-Kahlen wrote: >>>>>> >>>>>>>On April 14, 2000 at 12:46:43, Bertil Eklund wrote: >>>>>>> >>>>>>>>On April 14, 2000 at 10:57:27, Stefan Meyer-Kahlen wrote: >>>>>>>> >>>>>>>>>Yes, I have received about 75 games of Shredder but haven't checked them yet. >>>>>>>>>Some programs really have improved during the last months so I am busy working >>>>>>>>>on Shredder. This is my main target right now, if I have some time left I'll >>>>>>>>>check the games but I don't expect any problems. >>>>>>>>> >>>>>>>>>I give you a recent example of some autoplayer problems. This is part of an >>>>>>>>>email I wrote today. I didn't want to publish it first but it might help here. >>>>>>>>> >>>>>>>>>"You know about the autoplayer and all of its weaknesses, just this morning I >>>>>>>>>had a strange behaviour again. I was playing vs. X when X send the save game >>>>>>>>>command to Shredder 3 times in a row. The result was that Shredder saved the >>>>>>>>>last game 3 times and adjusted the display of the match result according to >>>>>>>>>that. Well, Shredder won, so he got 3 points instead of 1 for the game :-) If >>>>>>>>>you don't look very carefully at all the games played with the autoplayer you >>>>>>>>>can get all sorts of wrong results." >>>>>>>> >>>>>>>>Hi! >>>>>>>> >>>>>>>>As I have told you and mr Weiner several times, we partly follows most games and >>>>>>>>check the results of every game. In example Tony Hedlund is a very good player >>>>>>>>and played in the last Swedish correspondence Championship, Bo Aurell another >>>>>>>>tester played in the Swedish Championship in the sixties and I think they can >>>>>>>>see if there are any anomalies. I don't know what persons that test your >>>>>>>>programs, are they only reporting the scorenumbers? I have now played 122 >>>>>>>>tournament games with S4 with only two stops, one game where Hiarcs stopped >>>>>>>>within the opening book and one game where S4 played only one move (1.d4) and >>>>>>>>saved the game as a win for itself, Nimzo saved it as an unexpected end. >>>>>>>> >>>>>>>>So far I can't see what the problem is. S4 has won against Hiarcs and Nimzo and >>>>>>>>looks to crush Tiger (my games). Only one bad loss against F6 a program that >>>>>>>>many people believe is the best for the moment. >>>>>>>> >>>>>>>>If we had been allowed to test S4 when it was new it had probably been number >>>>>>>>one for some time. I believe S4 belongs in the group of Junior6 and Tiger that >>>>>>>>fights for the second place. It looks like Fritz6a have made a big step forward >>>>>>>>and I believe it is the best program for the moment. >>>>>>>> >>>>>>>>I also think that we have been fair against S4 allowing it to use the Turbo-cds, >>>>>>>>that seems to be much more efficient than the original Thomson-CDs. >>>>>>>> >>>>>>>>Bertil >>>>>>>> >>>>>>>>>End of quote. >>>>>>>>>I have just changed the name of the program involved as I don't want to blame >>>>>>>>>anybody. >>>>>>>>> >>>>>>>>>Stefan >>>>>>> >>>>>>>I didn't complain that you have treated Shredder unfair and also don't think so. >>>>>>>For me there is no problem. Also I didn't say storys like the one above happen >>>>>>>at the ssdf. I just wanted to point out that there is so much that can go wrong >>>>>>>using the autoplayer. >>>>>>> >>>>>>>Stefan >>>>>> >>>>>>I must disagree on this. I've played thousands of games with the autoplayer, and >>>>>>I would say that it's very, very little that goes wrong with it. >>>>>> >>>>>>Tony >>>>> >>>>> >>>>>I disagree with your disagreement. :) >>>>> >>>>>_Any_ piece of software that is dependent on timing delays and such is a piece >>>>>of trash. It turns the interface into a synchronous operation when it should >>>>>work asynchronously. trash for an interface. Compare it to the winboard >>>>>interface that has _no_ timing issues between an engine and the interface... >>>>>No way to lose a message, get multiple copies of the message, etc... >>>>> >>>>>maybe I insulted the trash of the world, when I think about it. :) >>>> >>>>hmm one gets the impression that Bob isn't very fond of autoplayer ;))) >>> >>> >>>How would you like a piece of software that requires the following: >>> >>>(1) a variable delay added so that if you want to move _instantly_ you have >>>to delay for some number of milliseconds or auto232 will miss the move and the >>>game hangs. >>> >>>(2) the delay has to change when in endgame databases as the probes cause a >>>large number of interrupts that change the timing in ways I never could predict, >>>which would again cause the game to hang. >>> >>>(3) gives the opponent the ability to tell you to 'save game' and such at any >>>time the opponent wants, even if it is _your_ move. >>> >>>(4) requires tampering with the delay when you change to a faster machine. Or >>>if you add more tablebases. Or if you remove some. >>> >>> >>>The communication protocol should be totally asynchronous, as in winboard. How >>>it was written to have these problems I wouldn't venture a guess... >> >>A techno-nightmare, or so it seems. >> >>But, ugly as it may be, it works just great and has been an immensely useful >>tool. >> >>Enrique > >Reading Bob's message, then yours, I find it hard to believe that it really has >worked "just great". :-) > >It's better than nothing, but there should be better than it. Winboard's >protocol also has its share of problems. I think that a replacement protocol >for controlling chess engines that ChessBase, Millennium, and independent >authors could all use is needed. > >Dave We talked about this once before, but no one other than 'freeware' guys were interested. It really is time to sit down and define a workable standard. Precise commands with non-ambiguous meanings, covering all the bases we want to cover, from different types of time control, to end-of-game messages, to how a game will be resumed after interruption, to you-name-it. I'm willing if anyone really wants to talk seriously (we could do it in a long thread here). But it would be nice to get commercial interest/input so that _everybody_ commits to supporting it... then it would work, just like PGN has worked.
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.