Author: Bertil Eklund
Date: 12:51:03 04/17/00
Go up one level in this thread
On April 17, 2000 at 08:21:25, Stefan Meyer-Kahlen wrote: > >I didn’t want to write this post first because I am getting so sick and tired of >the endless autoplayer discussion. I read two post in another forum which made >me pretty angry and upset and distracted me from my work so that I had to do >something. > >In these posts I was accused of being paranoid, having a bad faith and >manipulating to name a few. Of course I don’t like being called that way so I >think I have to explain my point of view again. > >First of all I don’t think that the ChessBase or any other autoplayer is >cheating by sending illegal commands or something similar via the serial cable >to disturb the opponent. I also don’t think that there is manipulation done on >purpose to favour one program over another. In several letters to me Millenium have accused Chessbase me and the SSDF for using a cheating auto-player. Mr Weiner quotes you as a source where you have seen a lot of "cheating" from the CB-autoplayer. I have asked him (copy Stefan M) several times to send me/us the evidence openly or anonymously, but so far I haven't received anything other than more accusations etc. I have received about a dozen letters from Millenium (Copy Steman M-K)including law-threats and all sorts of accusations and I haven't heard a single word from you where you said you disagree. >I hope that no author of any program or autoplayer felt personally attacked by >my statements, but if this were the case I apologize and apply to them to send >me an email to discuss this privately or here if he wishes. >I do think that there are so many problems with the autoplayer that it is very >easy to get wrong results in a match. To me it is mysterious how one can play a >long series of games without any problems. I also already said that this got >better recently but not so long ago this was impossible for me. Maybe it was my >fault and I was too stupid to set up the autoplayer correctly, but in every >match there were problems. > In example Enrique played a few houndred games in his tournament and had no problems at all, including CB programs. I think it is less manipulating and distortion in autoplayer games vs manually played games. >I just will give some examples. I won’t name any program but I think I have to >say that I had problems with every autoplayer, not only the one from ChessBase, >including mine. > >· Suddenly a program didn’t move any more. This happend in won, drawn or lost Netscape and Explorer hangs to. >positions. The game was adjourned with a random result. Most programs save the result according to the score including Shredder. +- a given score gives the result, if in doubt we outplays the game manually. >· Multiple save game commands were sent with the result that the automatic score >display and the database of the games got messed up. >· No save game command was sent with the same result as above. >· Programs crash, also in won, drawn or lost positions. >· Games were aborted for reasons unknown to me. >· etc. Mostly we play 2-4 games in a row so we can partly follow the games. No modern program loses on time in tournament time control. >So what happens for example to aborted games. First of all one has to notice >that there were problems. Therefore one has to check all the games in the >database and not only add the wins and draws and losses to calculate the >performance. So how many people are checking all the games of the autoplayer? >Even I don’t do that all of the times. And even if you are checking all of the >games manually, what are you doing if there are three identical games in a row. >Is it a multiple save or just the result of book learning on the winning side. > >Another issue is the autoplayer standard itself. For example there is no way to >control the time. It happens quite frequently on my machines that a program >loses on time, even on his own clock, and didn’t get punished for that. What >keeps me from using more time than allowed in an autoplayer match? > >Please don’t answer to this post that I shouldn’t complain so much and design a >new standard instead. > >The testers have to know about the autoplayer problems to try to avoid them. >This is my major concern and the reason why I am bringing this subject up here. >It happened more than one time before that one of the following occurred. > >· One program didn’t play with its optimal settings. You can argue that I can >restrict my program to forbid that, but then I will annoy my customers who want >to fiddle around with the engine options in Shredder. >· The same is true for opening books and endgame databases. >· Of course the two machines have to be identical or at least similar. >· The chess program must be the only running application. I thought at least >this must be clear so I was very surprised to learn recently that even that had >happened in the past before. > >Maybe all of the testers are more alert than I assume, but unfortunately reality >proves the opposite far too often. Dear testers, please don’t feel insulted now. > >I am thinking what to do in the future. The option to disable the autoplayer in >Shredder doesn’t seem like a perfect solution to me. > >OK, that’s it. If you still believe that I am starting a campaign or something >similar feel free to do so. If you still believe that I am paranoid, ok, but >keep it for yourself and please don’t insult me in a public forum. If you get >the impression that I am hiding because of stronger opponents, I will address >the tournaments I regularly join as often as possible. > >I can’t give you more arguments than that and also am not willing to spent all >of my time writing and reading postings in newsgroups. If you have your own, >different opinion on the whole issue I also will and can live with it. > >I am trying to make a living by writing chess programs. My sales and therefore >my income depends also on the outcome of tournaments and rating lists. What if >there is a random error margin involved in there? If I want random income I >rather go to a casino and play roulette. It is fine with me if one of my >competitors is better than me, also I really wouldn’t like that. I also have to >say that in contrast to what is written and assumed here I have, at least in my >opinion, a good relationship to all of my colleagues. > >Stefan I think it is better if we can follow the results from "independent" testers like Enrique, Djordje, mr Schumacher or the SSDF instead of the results from the Millenium-site were some anonymously AUTOPLAYED games shows Shredder4 winning over all it's competition in 15min games. There isn't anything said about the problems with autoplayer testing. Bertil
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.