Author: Robert Hyatt
Date: 20:47:41 12/13/03
Go up one level in this thread
On December 13, 2003 at 16:31:46, Sandro Necchi wrote: >On December 13, 2003 at 13:44:21, Robert Hyatt wrote: > >>On December 13, 2003 at 03:04:02, Sandro Necchi wrote: >> >>>On December 12, 2003 at 22:38:56, Robert Hyatt wrote: >>> >>>>On December 12, 2003 at 13:12:46, Sandro Necchi wrote: >>>> >>>>>On December 12, 2003 at 10:35:00, Anthony Cozzie wrote: >>>>> >>>>>>On December 11, 2003 at 13:20:29, Sandro Necchi wrote: >>>>>> >>>>>>>Robert, >>>>>>> >>>>>>>I think it is not the case to continuo. I will stay on my ideas as you are going >>>>>>>to stay on yours. >>>>>>> >>>>>>>I am interested on winning games on the board and not in the forum. >>>>>>> >>>>>>>I am sorry, but I do trust more Darse than you, as well as the TD in Graz. >>>>>>> >>>>>>>I only hope that in future the programmers will agree to stop the games when the >>>>>>>score is not lower than -10 to avoid "ridiculus". >>>>>>> >>>>>>>By being a chess player I find to continuo playing "extremely lost games" >>>>>>>offensive and not useful at all to show how strong the chess programs have >>>>>>>become. >>>>>>> >>>>>>>I am saying this here now to avoid someone would link this to Shredder games. >>>>>>> >>>>>>>I am a true chess and computer chess lover and hate to see non senses like >>>>>>>playing extremely lost positions. >>>>>>> >>>>>>>How can a programmer be proud of not losing or winning a game extremely lost? >>>>>>> >>>>>>>Does it makes sense a statement like "well, this year my program did score very >>>>>>>well as we scored 5 out of 8 while last year I scored 0. The first game it went >>>>>>>down -12, but the opponent had a bug and we could win the game. The second one >>>>>>>the opponent had a mate in 12, but a bug made the program lose 3 pieces and we >>>>>>>won. The third game we won with 3 pieces less because the opponent program got a >>>>>>>bug that removed all the hashtables use and so on..." >>>>>>> >>>>>>>Wow there is a lot to be proud! >>>>>>> >>>>>>>I am clearly exagerrating, but it seems for some people this would be >>>>>>>acceptable... >>>>>>> >>>>>>>??????????????????????? >>>>>>>I will never understand this! >>>>>>> >>>>>>>Sandro >>>>> >>>>>Hi, >>>>>> >>>>>>As a human, I get annoyed when people continue when they are down a rook or >>>>>>more. I get _really_ annoyed when they beat me anyway :) And I can see your >>>>>>point, its something of an insult: the other player is saying that they can win >>>>>>even though they have a horribly lost position. >>>>>> >>>>>>However, computer-computer games are different IMHO. Computers don't have egos. >>>>>> They never get tired. Why not let it go all the way to checkmate? >>>>> >>>>>I was not referring to 2003 WCCC, but I was proposing something for the next >>>>>tournaments. >>>>> >>>>>My point is: >>>>> >>>>>1. Since the programs now are much stronger than 20 years ago, why not change >>>>>the rule about resigning and let them resing when they are down -10? >>>> >>>>I don't understand the request. >>> >>>The request is simple: >>> >>>If we do not want to see the computer tournaments as a private affair for the >>>programmers and look to get involved more people to watch the games and make >>>them more interested/fun about these events, then we should try to understand >>>what they think and what they like to see. >>>What I know is that they would like to see the programs play more likely to >>>human players and therefore resign hopeless positions. >>>Which are hopeless positions? >>>To me when you are a piece down, but since this would leave to many chances for >>>recovering the disadvantage, than I think we better increase that to a higher >>>level leaving chances nearly to 0. So than -10, which is equal to 2 rooks or a >>>queen down seems more reasonable. >>>This is where I make the statement. >> >>I _still_ don't understand the request. > >The request is: > >when there is a tournament with live games (games that people which are not in >the tournament hall can see), then FORCE every program to resign when they reach >-10 to avoid showing boring parts of the game where the outcome is sure 999 >times out of 1000. I would not object if FIDE were to adopt that rule. I once watched a master try to beat an IM in an ending that was theoretically winnable, but the master obviously had no idea how to do it, and the IM kept "escaping". But they played to the 50 move rule before the draw was official, as the IM could not force a draw and the master would not accept/offer a draw. Technically, it was his right according to the rules. I would much prefer to see ICGA events use automatic interfaces. Then there is _never_ any need to resign. Let the programs play to mate. > >If there are not live games, then let the programs resign when they want (the >team wants with the TD operator approval. > >>If the _computer_ says "I resign" >>then the game ends. And you can set the resignation threshold to any value >>you want. > >OK, but my request is change the rules to force everybody to resign on -10 >without cheating. > >>The ICGA rules simply make the _operator_ ask the TD if the >>_operator_ wants to resign for the program. But if the program resigns itself, >>there is no discussion at all, the game simply ends. It has always been this >>way... If the program can offer or accept draws, no TD action is needed. > >I agree, but it should be made in a clear way that nobody cannot understand >otherways. Your latest one is OK! > >>The >>TD only prevents the _operator_ from resigning, or offering/accepting a draw >>outside of the program's frame of reference. > >I agree. This is why I am asking to change it in a way which will be clear for >everybody. > >>The program can do whatever it wants, whenever it wants. And the TD can't do >>a thing about it. IE I can say "I resign" and if the TD doesn't like it, >>fine. But I'm not moving. So we sit until the flag falls. > >I agree, unless the score is less than -10 and it has been agreed to change the >rule. In that case the program should resign and nobody should complain. > >>That's why I don't understand your request, because it is already within the >>rules if the _program_ handles it. Mine does... > >I am saying that ALL programs should be made the same way as yours. > >>> >>>I am not saying just because I am requesting this everybody must agree. Mine is >>>only a request to improve this field by making more people more attracted to it >>>and not only computer chess lovers. >>>I only ask to think about this. >>>I have nothing to gain in this. I do it only because as I said I am a true lover >>>of chess and computer chess. >>>This is the reason why I have spent to much time and money in this field. >>>You can say I do not agree. It's OK, but you cannot say it is not worth to think >>>over it, I guess. >> >>Note that I already resign in appropriate cases. > >I was not saying something different. I was talking in general and not referrign >to your program. > >>However, remember that >>I have probably watched 100X more computer vs computer games than you have, > >It is possible. I think I have seen between 50,000 to 60,000. How many did you >see? I have no idea. I have watched them every night for 8+ years, plus off and on during the day and on weekends... > >>thanks to my ICC presence for 8+ years. I have seen my program draw from >>+9 when it missed a way deep perpetual. So in comp vs comp games, I don't >>resign, because in the middle of a game I can't say "hmm. this is not a >>queen-type ending where perpetuals are common, so I want to resign at +4, >>while when queens are on I would prefer to wait until +10 before I give >>up. But we don't allow on-the-fly adjustments, so I can't do that. And I'd >>rather err on the side of safety. IE if you get to +9 and fail to find a >>perpetual, that really is your bug for missing it, not mine for playing on >>to allow you to miss it... > >OK, of course a bug can change the outcome, but how many chances there are today >of such a bug in that specific case? >If there is 1 out of 1000, is it worth to do it? The point is that the cost of doing this is zero in a good event. No humans have to type, the computers do it all automatically. And since there is no "pain" let 'em play. >My answer is no. >Pls. remember that today programs are much stronger than the ones in the 70' and >this reduces the chances to recover a big disdvantage. > >>>>A program has _always_ been able to resign >>>>on its own, at any point it chooses. The operator is more limited in what >>>>he can do. But if a program says "I resign" then the TD has always accepted >>>>that at any event I have played in. > >Exactly. This is what I am saying and let's set this at -10 which is a clear >advantage for the opponent. Clear yes, but I have seen -10's turn to draws. Not often, but not zero either. > >>If I wanted to resign for my program (I >>>>have not had to do that since mine has self-resigned for years) I had to clear >>>>it with the TD. But not if the program made the choice. >>>> >>>>However, it seems you want to _force_ this to be the policy, >>> >>>Yes, this is the idea to make improvements. This is my opinion of course and I >>>do believe many people would agree with it. I am proposing something before the >>>tournament starts, to make it the same to everybody. >>> >> >> >>The problem with forcing a resignation threshold is it _must_ be a safe >>one. I can provide games where either my program or the opponent was at >>+9 and drew by missing a very deep perpetual or one of "those" stalemates >>that are so hard to see. > >Well, first of all I am speking of +10 (-10) and considering the stronger >programs there are today I think this is a nearly safe value. of course it is >not 100% sure, but very close to that. > >>>>and I don't agree >>> >>>OK, you do not have to agree. It is up to you to do it. Simply think why I am >>>asking this, before you decide. >>> >> >>I understand your request. But notice my response. I want the _programs_ >>to win/lose/draw the game, not some hokey rule that has a critical exception >>nobody thought about. > >Nothing has only advantages or disadvantages. I think the disadvantage is that >you would leave a quite rare possibility not to lose a game or to win it, but as >an exchange the games will be more interesting to chess players who would like >to see the computers play more like the strong human players. > >It is an idea for improvements. If we alway keep things like they are we may not >improve anything... > >>>>with that, particularly with sudden-death time controls. >>> >>>OK, than at blitz, even if I think that that would be good as well we could >>>leave it as it is; I mean up to the mate. >>> >>>> >>>>>2. It is true that a bug may help the program which is lost, but which are the >>>>>chances today? Is it correct to say 1 every 1000? If this is true, why not >>>>>concentrate to improve their play on the first part of the game rather then >>>>>hoping to be extremely lucky in the endgame? >>>>> >>>>>Yes, it is true that they do not get tired, but the people watching these games >>>>>do and they would switch to another game as that is of no interest anymore when >>>>>the advantage is so high. >>>>>This is what I do and I do believe I am not the only one. >>>>>I guess we all want to have more people attracted by chess and chess programs, >>>>>so why not give them something they would prefer? >>>>> >>>>>This is only a proposal for the next tournaments, to make them more attractive >>>>>for the real chess players. >>>>> >>>>>Sandro >>>>> >>>>>>Do you >>>>>>think you deserve to win if your program can't play a simple mate in 8? >>> >>>I thin we deserve the win if the opponent is not able to beat us and or to ask a >>>draw. >> >>I disagree. In 1970, at the first computer chess event ever held, I watched >>a game between Coko and Genie, where Coko did not discern between mate in N >>and mate in N+1, and kept playing a mate in 2 move every time it had to move. >>It eventually lost the game. You not only have to reach a won position, you >>have to _win_ it as well. That is the point for the accepted rules of chess. >>FIDE, you might notice, does _not_ have a forced resignation threshold in >>their rules... > >OK, but now we are in 2003. I think in 33 years things have changed a lot so the >programs are much stronger and the hardware is quite faster... I played a commercial program a few months ago and was a rook and piece plus a couple of pawns down. But the score kept climbing until it finally hit zero. But the checks could not be stopped, although the king could reach nearly every square on the board making detecting the repetition _very_ difficult. So difficult neither program saw it for a good while, but it eventually was drawn... Takw queens off, the deep draws are less rare, unless there is a possiblity that the weaker side can walk his king into a position where it has no legal moves, and then uses its remaining rook to check the opponent and he can't take it without turning it into a stalemate. Both of those cases happen. > >>>>>> >>>>>>anthony >>> >Sandro
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.