Author: Robert Hyatt
Date: 07:28:27 12/25/03
Go up one level in this thread
On December 25, 2003 at 03:27:02, Mike S. wrote: >On December 24, 2003 at 21:41:09, Rolf Tueschen wrote: > >>>>>>(...) > >>NOPE Mike! That a 3-fold, once you notice it, is a draw, this is NOT something >>special for computerchess, but for human chess and from there it was taken to >>computerchess. > >You certainly know (?) that a 3-fold is *not* automatically a draw once one >notices it, but *only when claimed* It doesn't have to be a draw. The same goes >for the 50 move rule. The only "must" draw like that is stalemate :-) Here is the question you have to answer: "what engine will "notice and inform you about a 3-fold repetition" but not intend to claim it? The answer is none. If an engine repeats for the third time, it does so with the full intent to see the game end, because computers don't implement this as a "I will repeat, but hope my opponent won't notice and claim it himself." The search doesn't work like that, and no program I know of does so either. I don't believe a TD would take that definition either. IE if he walks by and a player says "3-fold repetition" and the TD looks and it is correct, he would record the result and move on. He would not ask "you said 3-fold repetition, are you just pointing it out, or are you claiming it?" Because the FIDE rules don't distinguish between the two, nor do FIDE or ICGA rules specify specific wording for claiming a 3-fold repetition. Offering or accepting/declining a draw is a different matter, and at least my program is capable of offering and accepting with no human interference. There is a well-defined protocol for offering draws and accepting them, dictating how the offer should be made and accepted. But for claiming a 3-fold repetition, or a 50-move rule draw, there is no prescribed wording that is required, meaning that an obnoxious pop-up saying "3-fold repetition position" <OK> is more than enough since the player has to do _something_ to continue the game or the program will sit there and not play further. That seems to imply, at least to me, that this was not a simple "informational thing" because you don't have to click OK on information such as PVs and scores, before the game continues. But here he did. That makes it "different". > >When I'm (otherwise) *winning*, i.e. a rook up etc., and my opponent forces a >perpetual, I won't claim a draw of course. Still, I may take notice of it. But >it makes no sense to claim. Also, I *don't have to claim* (of course independent >from who's better). If the opponent misses it - his business. That's how it >actually really is in chess in general (I don't quite know what else you are >talking about). > >I guess you may assume, all computerchess programs would be so foolish to claim >the draw against themselves then in a situation like above, but that's not true >either. I definetely know Nimzo 2000 (own GUI) doesn't claim such draws when 0.5 >would be lost by it so to speak. I guess Fritz in serious rating game mode will >behave the same, but I didn't test that. Since the GUI is handling this, I don't see why the engine would matter. But in the case of Fritz, we _are_ talking about the Fritz GUI don't remember, and it did make the claim in the game in question. > >I'm not talking about a program which has the bad position and is happy to draw >(not about the Jonny case anymore, but more general), but I'm talking about the >program who has the much *better* position, but couldn't avoid the repetition. >Now, how does this program *only inform* the operator that is has spotted the >repetition, but not claim at the same time? That is a circular and pointless argument. IE the program repeated for the second time and couldn't avoid it. If it can't avoid the 3rd repetition either, then what possible point is there to continue, since it also won't be able to avoid the 4th, or the 5th, if it couldn't avoid the 3rd. This seems to be based on a humans ability to overlook the obvious (the repetition) and give you another chance. Computers won't overlook anything, and if you think you have winning chances, you vary before the 3rd repetition to keep the game alive. That is the _only_ way to play. And it is certainly the way the tree search handles this... > >I think you've really misunderstood that. > >So, regarding your view of the "two sentences", the program would have no other >option than to remain silent. I see no other option currently, when the rules >(written or not) are so ambigous. Imagine you're the opponent operator with an >engine in early developement, repetition rule not implemented, you're a rook >down but can force a repetition (engines will do that no matter if the rule is >implemented or not, because it keeps the opponent from increasing his >advantage). Now, when the *other* program just pops up "3-fold repetition" meant >as an info only, you will run to the TD claiming your opponent's program has >claimed a draw, but the evil operator continues to play... :-)) How can it be "information only" when you have to click OK to dismiss the window? Do you have to click to see the clock count down? Click to see the next PV or score displayed? Click to see the next book move played? When the computer program stops the game until the operator takes an action, that is done to get the operator's attention. That doesn't sound like an "informational thing" to me at all. Such would require _no_ operator action so that the game is not delayed or lost on time. > >That's why this has to be clarified in a way, that the info and the claim can be >distinguished from each other. Except that there is no such "info" that comes out of a chess program. :) > >(I did my best to explain :-)) > >>Writers who imply that it is justified that the party with the worse position >>should NOT claim a draw in case of a 3-fold must repeat classes and enter the >>special lecture of Geheimrat Rolf about logic and chess history. The loop >>remains until writers admit that they were absolutely wrong in their perception. > >Of course they should claim the draw :-) and every programmer will certainly >care that they do. I just say when they claim, they should have to use the words >"draw claim" or "I claim draw," but just "3-fold repetition" is not sufficient >IMO, as explained above. Also, think of cases where it's not so clear which >position is better or worse (which is not so uncommon). The program must be able >to give at least an information (only) that a repetition happened in a way that >it's not a draw claim by any rules, at the same time. > >I don't agree to an opinion, that a 3-fold in a computerchess game *must* >inavoidably and immediatly end the game as a draw. This is not chess culture, >and the FIDE rules are different. It must be claimed. No claim, no draw. > >OTOH, in the general or private practise (of automatic engine tournaments), in >contrast to "official" tournaments, this topic doesn't exist anyway. Because >then, this is handled by the software alone. When a repetition occurs, it will >be always a draw, in an engine match or Auto232 match I think. I wonder if there >are exceptions, i.e. when both engines evaluated themselves better after an >"unforced" repetition, not by perpetual... There's not search function for that. >Very unlikely, but probably not impossible. > >More common are accidents like engine 1 misevaluating a 50 move draw in the 99th >ply and making a stupid move which allowes engine 2 to mate immediatly in the >next move, because the mating move is neiter a pawn move nor a capture. I seem >to remember a report, that a program (or a chess computer?) played the mating >move and claimed a 50 move draw :-)) So if the claim comes (correctly) before >moving in such a situation, but it's visible that the program actually has and >plans a # in 1 at the same time, lets just hope that this will never happen in a >decisive game in the last round of an ICGA tournament :-)) > >>BTW I find it extremely sympathetic that in special Chr. Liebert from CSS has >>defended the LIST author with so much energy. >> >>I can only hope that Fritz gets a fair compensation if it should be proven that >>the reproaches against his engine were false. Here: couldn't be proven at all. >>Also I find that the lack of solidarity by all the other programmers is crying >>loud to Heaven! > >About that, we both (and others who usually disagree :-) are of the same, or at >least similar, opinion, so we must be right... IMO the protest wasn't founded >well enough to justify the demand to provide the source code. I don't assume >that there are exactly defined standards regarding the requirements for a >protest to be valid, which then could lead to that in consequence. I'd say there >should be somewhat high standards (strong evidence) be required, to be able to >lead to a disqualification in case the source code is not provided, in respect >of the "in dubio pro reo" principle. Not providing the sources in itself is not >such a big crime, so if there's no proof for the clone accusation, I don't >understand such a long ban either. > >But I'm afraid, that case has been closed already. Maybe we'll see after the >next scandal, if the rules have been improved then. Every amateur should be >ready to face such an accusation, and take steps respectively. > >Form my viewpoint, if I was a chess programmer, I'd not want to join a >competition even with the risk to face such an accusation out of the blue, under >such insufficient circumstances. Even that is hardly tolerable. Probably it was >easier to defend oneself in front of the holy inquisition (ore they'd simply >burn him as well, when he doesn't provide his secret book of wizard spells). > >Meanwhile, List is in continueing developement and performing very fine (better >than some pros). > >>All the best to you, Mike, and enjoyable holidays and a Happy New Year, >> >>Rolf > >Same to you, >Mike Scheidl
This page took 0.01 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.