Author: Mark Young
Date: 06:12:05 11/17/98
Go up one level in this thread
On November 17, 1998 at 08:42:21, blass uri wrote: > >On November 17, 1998 at 08:11:45, Mark Young wrote: > >>On November 17, 1998 at 07:07:55, Albert Silver wrote: >> >>>On November 17, 1998 at 05:05:05, Mark Young wrote: >>> >>>>On November 16, 1998 at 22:02:29, Christophe Theron wrote: >>>> >>>>>On November 16, 1998 at 20:22:03, Mark Young wrote: >>>>> >>>>>>On November 16, 1998 at 18:32:16, Christophe Theron wrote: >>>>>> >>>>>>>On November 16, 1998 at 17:46:32, Mark Young wrote: >>>>>>> >>>>>(snip) >>>>>>> >>>>>>>Thorsten did not say Junior (Amir) is cheating. >>>>>> >>>>>>You need to go back and read what Thorsten wrote, because it is clear to me that >>>>>>what he is saying. >>>>> >>>>> >>>>>Here is what Thorsten has said. BTW this was a reply to one of your post, so you >>>>>should already know: >>>>> >>>>>-----------(Begining of Thorsten Czub quote) >>>>> >>>>>amir is not doing the GUI. don't you get this ? >>>>>amir has only send his engine to chessBase. he does not know what they do with >>>>>it. >>>>>where am i saying amir cheats ? >>>>>it is the engine X in the user-interface. if the user-interface is buggy, ANY >>>>>engine would make problems. >>>>> >>>>>how do you want to construct now that i call amir a cheater ? >>>>>imo you only want to make trouble here. you mix things. you claim things >>>>>i have never said and would never say, because i don't think it is amir who has >>>>>done that. >>>>>this was all done long before amir. >>>>> >>>>>-----------(end of quote) >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>>>>>And because Junior 5's autoplayer has >>>>>>>>a bug this proves that Fritz 5 is doing the something and proves that Fritz 5 >>>>>>>>also cheated. >>>>>>> >>>>>>> >>>>>>>F5 autoplayer has been reported to have the same kind of bug several months ago. >>>>>> >>>>>>Really, and now we know the that report was wrong, Ed checked it out and found >>>>>>no problems with it. >>>>> >>>>> >>>>>To my knowledge, Ed did not test the autoplayer, because he never had it. >>>>> >>>>>Ed has been told that Rebel has sometimes been prevented from saving its games. >>>>> >>>>>He tried to figure out if this could have hurt Rebel's learning feature. >>>>> >>>>>He thinks that it has not hurt Rebel's learning algorithm. >>>>> >>>>>That's all. I hope Ed can confirm this. If I am wrong, I have to be corrected. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>>>>That is: somebody has found that sometimes F5 opponent was unable to save its >>>>>>>game. This "somebody" was not Thorsten. >>>>>> >>>>>>I can only go by what Ed reported. He is the only expert I know that has seen >>>>>>it. >>>>> >>>>> >>>>>I don't think he has seen the bug himself. Once again I hope he will confirm. >>>>> >>>>> >>>>> >>>>> >>>>>>>>Known facts >>>>>>>> >>>>>>>>1. Junior 5 has a bug in the autoplayer. >>>>>>>> >>>>>>>>2. Fritz 5 and Junior 5 have different autoplayers. >>>>>>>> >>>>>>>>3. SSDF is not having this problem with the Junior 5 autoplayer. >>>>>>>> >>>>>>>>4. Ed said that the Fritz 5 autoplayer was "clean" >>>>>>> >>>>>>> >>>>>>>I guess you are going to have a message from Ed for this, because he did not say >>>>>>>that. >>>>>> >>>>>>I think that what I read. And I read it was "clean" But if I am wrong Ed can >>>>>>correct me. I want the facts to be correct. >>>>> >>>>> >>>>>Once again I hope Ed replies to this. >>>>> >>>>>In between, may I quote one of Amir's post: >>>>> >>>>> >>>>>-------------(Begining of quote) >>>>> >>>>>It's not yet clear what this J5 autoplayer problem is, and whether we are >>>>>dealing with a single problem or several unrelated ones, but I don't understand >>>>>what this has with saving the opponent's game. In the past, ChessBase >>>>>acknowledged a problem in the F5 autoplayer's saving of opponent's game, but at >>>>>least apparently this is working ok in the J5 autoplayer. >>>>> >>>>>-------------(End of quote) >>>>> >>>>> >>>>>I know Amir is not talking for ChessBase, but he is in close relation with them. >>>>>So we have at least another confirmation that there has been "a problem in the >>>>>F5 autoplayer's saving of opponent's game" (Amir's words). >>>>> >>>>>Of course, ChessBase people are free to correct this if they want. If they don't >>>>>know CCC, maybe somebody can tell them we are talking about them here... >>>>> >>>>> >>>>> >>>>>(snip) >>>>> >>>>> >>>>>>>>2. How can anyone draw the conclusion that since Junior 5's autoplayer has >>>>>>>>somekind of problem on some computers, this shows that Fritz 5's autoplayer was >>>>>>>>doing the same thing? >>>>>>> >>>>>>> >>>>>>>Nobody needs to make this conclusion. >>>>>> >>>>>>> >>>>>>>The problem with Fritz5 was reported several months ago by somebody else that >>>>>>>has tested the Fritz5 autoplayer. >>>>>>> >>>>>>>But by reading your sentence suddenly I realize that both autoplayers come from >>>>>>>the same company. Thanks for the idea. :) >>>>>> >>>>>>I can see you are as loose with the facts as Thorsten, >>>>> >>>>> >>>>>Did I say anything that is wrong ? >>>>> >>>>>Maybe you didn't know that F5 autoplayer had a problem ? >>>>> >>>>>Maybe you don't know both autoplayers come from the same company ? >>>>> >>>>>Where am I loose with the facts ? >>>>> >>>>> >>>>> >>>>>>What a reckless statement >>>>>>to make. By reading your sentence I suddenly realize that fair play and a honest >>>>>>discussion about other programs means nothing to you. I guess for some it is >>>>>>worth being dishonest, if they think it can give their arguement more sting. >>>>>> >>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>>>3. How does anyone draw the conclusion that because a piece of software has a >>>>>>>>bug, this proves intent that it was no mistake, just because a bug exist. >>>>>>> >>>>>>> >>>>>>>In the case of Junior, Amir's behaviour shows clearly that there is no >>>>>>>intention. >>>>>> >>>>>>Correct I agree. >>>>>>> >>>>>>>But there is a question because a previous secret autoplayer from the same >>>>>>>company had a related strange behaviour. >>>>>> >>>>>>How do you know it has the same strange behaviour, when it is secret. You can >>>>>>not base this conclusion on any kind of facts. >>>>> >>>>> >>>>>Once again, Amir said: >>>>> >>>>>-------------(Begining of quote) >>>>> >>>>>It's not yet clear what this J5 autoplayer problem is, and whether we are >>>>>dealing with a single problem or several unrelated ones, but I don't understand >>>>>what this has with saving the opponent's game. In the past, ChessBase >>>>>acknowledged a problem in the F5 autoplayer's saving of opponent's game, but at >>>>>least apparently this is working ok in the J5 autoplayer. >>>>> >>>>>-------------(End of quote) >>>>> >>>>>you are right, these are not facts. >>>>> >>>>>This is just what somebody (Amir in this case) says. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>>>>When you don't want people to suspect you of doing something nasty, you just >>>>>>>release a public autoplayer. >>>>>> >>>>>>And if you don't release a public autoplayer it gives people like Thorsten and I >>>>>>guess you the right to make up anything you want about the company and the >>>>>>autoplay. Because you did not like the fact that it was not public. >>>>> >>>>> >>>>>For months I did not say anything about this secret autoplayer. >>>>> >>>>>I thought: these guys out there are making too much noise against ChessBase >>>>>without giving evidence. >>>>> >>>>>So I can understand your point of view. >>>>> >>>>>I did not know that F5 had a problem with saving opponent's game. >>>>> >>>>>Now we know and we have a confirmation from at least 2 people (the guy that >>>>>originally reported the F5 problem and Amir). >>>>> >>>>>Thorsten has experienced the problem when playing Tiger against Junior, and >>>>>reported to me. Maybe it is not the same autoplayer, but this made me realize >>>>>that it was possible. >>>>> >>>>>I did not know, or did not believe, that the opponent could change the >>>>>autoplayer behaviour to the point that you could not even save your game! >>>>> >>>>>I am using the standard "NONAME" autoplayer. That means that on the computer >>>>>running Tiger there is a small piece of software, written several years ago by >>>>>C. Donninger, that takes care of communications with the opponent and sending >>>>>keystrokes to Tiger. >>>>> >>>>>I thought that saving the game was done automatically by this piece of software >>>>>when the game ends. The game ends with a timeout (no more move are played for a >>>>>given amount of time), so the opponent has no control to avoid this. >>>>> >>>>>What I did not know is that the opponent sends the command to save the game to >>>>>my own program. And that a bug in the opponent could prevent my own program to >>>>>save the game, and to learn if my learning algorithm is done when I save the >>>>>game! >>>>> >>>>>So an autoplayer is really a critical piece of software, and not releasing it >>>>>publicly is a problem to me. >>>>> >>>>> >>>>> >>>>> >>>>>>I'm sorry I have I problem with those kind of ethics. >>>>> >>>>> >>>>>I have an even bigger problem with the fact that the autoplayer is secret and >>>>>that the SSDF trusted it enough to establish results on the strength of the >>>>>chess programs. >>>>> >>>>>I have been a little bit late to come to this conclusion, but now I think it >>>>>would be fair that previous results of Fritz5 are removed from the list, and >>>>>that Fritz5 is rated with the new public autoplayer. >>>>> >>>>>I know it would require a lot of additional work from the SSDF guys, and I am >>>>>sorry for them, but it would definitely clean them from all suspicion. >>>>> >>>>>Do you think it would be unfair ? >>>> >>>>Yes this is unfair, I have seen no proof that there is anything wrong with the >>>>autoplayer. Just smear, because you guys did not like the fact that the >>>>autoplayer was not public. No, this kind of tactic can not stand. If SSDF falls >>>>to this kind of tactic, then SSDF is no longer independent. >>> >>>I don't know about that. I think they lost their independance the minute they >>>allowed someone's program to dictate special conditions. >> >>For this statement to be correct, then SSDF was forced to test Fritz 5 by >>Chessbase. I don't think the "special conditions" were out of bonds. >> >> It's true, that >>>increasing the RAM was then passed on to all other programs, but what of it. >>>Suppose Ed put in a command that only allowed his program to be tested on an AMD >> >>Well if Ed buys them all AMD machines I'm sure they would be glad to test it >>that way. But can Ed force SSDF to test this way, no. That is SSDF's call, and >>no one elses >> >> >> >>>processor, a processor that Rebel is known to favour? Sure, you could then say >>>that all programs will be allowed to run on AMD processors to equalize the >>>score, but I think fairness went out the window in this story a long time ago. >> >>I would not mind if SSDF would test on AMD machines, because many programs other >>then Rebel like AMD machines better, MHZ for MHZ. >> >>I would mind if Chessbase started to smear Rebel because the programs are on AMD >>Machines, and they use this as an excuse for their poor showing. >> >>You have to understand the conditions of the testing, example we do not know if >>Fritz 5 is best on P II 400 computes vs all the other programs. > >We do not know even if Fritz5 is best on pentium200MMX vs all the other >programs. We do know it is the best on the SSDF list. > >I found that Rebel9(p90) lost the same game some times against Fritz5 in the >ssdf games. >Rebel9 is a program that can learn so the only reason that I can find for it is >that the learner of Rebel9 was disabled by a bug in the fritz5 Autoplayer It could be but we don't know. It could be human error, or something else. (Ed can >correct me if there can be another reason). > >I found also that in the ssdf games Fritz5 won the same game 5 times against >Rebel8. I don't like repete wins, but they count on SSDF, and they have always counted. > >If fritz5 played less games against programs without learning feature the >results could be different And if my grandma had wheels she would be a wagon. > >I do not think that repeating winning lines is productive against humans in a >match because the human can predict the opponent of fritz5 and prepare against >it. Maybe true, maybe not, but we are talking about computer vs computer chess. >It is better(against humans) to learn by avoiding losing lines and not repeating >winning lines. You don't know that this is better, it maybe better it may not be. But Fritz 5 may not repete it the next game or many games. If does not work that way, it has other lines it could and would play. It just has a greater chance that the line would be played again, if it wins. And if it loses it has a much less chance to be played. It does both from what I can tell. Only if the same line loses many times will it not play that line again. And it will only keep repeting the same line over and over again if it keeps winning with it every time. A very good set up for playing humans and computers IMO. > >Uri > > > > > >> >>I'm finding Junior 5 better then Fritz 5 on faster computers, but I know the >>results could be different on slower computers. So I am not upset that Junior 5 >>is not #1 on the SSDF list. >> >>P.S. I think Junior 5 will be #1 on the SSDF list in its next posting but it may >>not be. >> >> >>> >>>Another thing, which has nothing to do with the above: did SSDF have any >>>problems with saved games with F5? If they had none (they might have had some >>>but assumed it was a glitch and not reported it), fine, but if they did then it >>>should be further investigated for Amir said that when he was able to spot a bug >>>it only happened when his program was lost. Not pointing fingers here as I don't >>>think cheating was involved, but this would affect the results. >>> >>>> >>>>And again you can not link the Fritz 5 autoplay and the Junior 5 autoplayer, We >>>>only know about the Junior 5 autoplayer. And SSDF is having no problems with it. >>>> >>>>Even under you own logic that they are linked with the same bug, SSDF is not >>>>have problems with the Junior 5 autoplayer, so under you logic SSDF should not >>>>have had a problem with the Fritz 5 autoplayer. >>>> >>>>So get over it. The new programs are comming that will be better then fritz 5, >>>>and they have public autoplayers. >>>> >>>>> >>>>> >>>>>(snip) >>>>> >>>>> >>>>> >>>>> Christophe
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.