Author: Jeroen Noomen
Date: 09:57:52 12/29/97
Go up one level in this thread
On December 29, 1997 at 06:37:08, Chris Whittington wrote: Dear Chris, I read your comments on Marty's posting. I have the feeling that you also have your doubts on what is going on! Furthermore I talked to Sandro Necchi a few years ago at the AEGON Tournament. On my question why MCP did so well in testmatches (he stated MCP would beat Genius 3 by a margin of 90%), he simply replied 'because we are outbooking Genius'. What do you think of that!? > >On December 28, 1997 at 13:11:06, Jeroen Noomen wrote: > >> >>The only two things I can conclude are following: >> >>1. Chris Whittington says it is normal that computerbooks are full of >> nonsense variations of 37 moves, ending with +4 or more. Variations >> that can be found nowhere in openingbooks or whatever. This is >>morally >> acceptable and people who don't agree with this, are immoral. > >1. I don't say 'normal', I say its data, and therefore possible to do >it. Of course it is possible. But don't you agree that the more a program makes use of such 'data', the more it is depending on HUMAN interfering? Instead of using its own playing capacity? >2. I don;t call these variations 'nonsense'. If Mchess (or anything) can >play them out and Rebel (or anythign) will respond by playings its part >to follow the 'nonsense'; then 'nonsense' can't be an appropriate >description, can it ? I still think that 40 moves of opening lines that have absolutely nothing to do with theory, a played game, an analysis by a player, or presenting a forceful line in a known opening variation is 'rubbish'. Show such a variation to an IGM and I'm sure he'll say something like 'what the hell is this?'. >3. Who said an opening has to be found in an opening book ? Openings are >subject to change, development, progress. (a) must programs never have >anything that hasn't found its way into the general knowledge yet ? (b) >can programs not develop their own kind of openign knowledge ? Of course not. I don't state this. But do you think it is normal that a line of 80, 90 plies that can be found nowhere and is 40 plies AHEAD of existing theory, is contributing to theory? Sorry! I don't think any strong player would agree on that one. > >4. If its raw data its presumably neither moral nor immoral. It human interfering to make the results better than they normally would be. >5. I've claimed neutral moral status for opening books. Right. Let the programs play it out themselves. They are strong enough, all of them. We don't need such tricks to make people sure that they are strong. >6. I did not say you were immoral. Read the very last line in this whole thread. >>2. When Chris Whittington plays the WCCCC 1998 against Bonecrusher X and >> this one plays 40 moves out of book against CSTAL, leaving the book >> and announcing mate in five, Chris says 'well, bad luck. Bonecrusher >> was simply stronger/cleverer than CSTAL'. > >No, I'll say what a pillock I was not to have covered this possibility. >Then I might come and offer Jeroom large sums of money to work on my >opening book :) > You cannot cover EVERYTHING. BTW, what is your offer!? :)))) >> >>For somebody who is an adherent of A.I., the statement that cooked lines >>are acceptable, moral and normal is rather strange. > >1. I say they are data and morally neutral. They are not neutral IMO. >2. I say that any data is acceptable. >3. Normal means what ? Did I say it was normal ? You don't object to such data. > >>I thought you were >>strongly in favour of programs containing CHESS KNOWLEGDE? > >I'm not so much in favour of it, as seeing it as necessary to play chess >:) Well, did you change your opinion on that one!? I always thought CSTAL is a B-strategy program.... I read some very interesting articles about CSTAL beating Genius because of pure knowledge. >>Can you tell >>me >>what such openinglines have to do with STRENGTH/KNOWLEGDE of a program? > >About the same as anythign else. Endgame databases or a hypothetical >machine that stored all the moves played out to termination. Before the endgame and middlegame there is an opening. If all programs use the same databeses, nobody has an advantage or disadvantage. If one or two programs use killer lines, the chances are not fair. Because the others do NOT have access to those lines. Besides, after a killer line the endgame tablebases are useless, because you'll never reach the endgame..... >Isn't the problem that you (we, all of us) have elevated the SSDF list >to some kind of oracle / god-like status. Its supposed to be *right* and >*accurate* and and and and moral probably as well. So anything which >'perverts' the list is the anti-christ; and must be opposed. >IMO we'ld be better off judging chess programs by qualitative, human >considerations, and not relying on these quantitative lists. But where can a chess enthousiast find such information? If he wants a program, he/she watches an ad, or the rating list..... >Also IMO you (we) are all racked by this silly competitive pecking-order >status-seeking nonsense which leads us to ........ well, insert your own >favourite twaddle here. >What are our programs supposed to do ? IMO they are games for >entertainment for humans; games should be fun and interesting to play. >IMO (again) you (we) have lost sight of this; instead being wrapped up >in pecking orders, lists, data morality and and and. >>>> >>>>Sorry Chris, this is nonsense. >>> >>>No, its an argued case that you don;t agree with. >>> >>>> >>>>37 moves of 'theory' and leaving the book with +4 has nothing to do with >>>>auto-learnt. It has also nothing to do with holes in an opening book. >>>>If you want to defend this: fine. This says a lot more about you than >>>>about me. >>> >>>Yabba dabba doo :) >>> >>>> >>>>Furthermore I once more give a brief overview about what happened: >>>> >>>>1. Thorsten Czub finds out that in a game MCP-Hiarcs the former one >>>>plays >>>> 37 moves out of book ending with u huge score in favour of White. >>>>2. Thorsten states that he doesn't like this kind of stuff. >>>>3. I confirm his findings, stating that I have seen this all before. >>>>4. I also published my view, that computers should stick to normal >>>>theory >>>> and battle it out THEMSELVES. Instead there seems to be a development >>>> that the battle between programs is fought in the rooms of >>>>openingbook >>>> makers. >>>> >>>>And THAT is a development I don't like. Whatever you think, you will not >>>>change my opinion on that. Maybe the Nunn-method - a preselected set of >>>>opening positions - is a good solution for this. >>>> >>>>If you'll find out one day that the CSTAL book is attacked by other >>>>programs by using a lot of nonsense lines that have nothing to do with >>>>opening theory and you'll need weeks to 'repair' this, maybe than you'll >>>>understand me. And I am sure that at that moment you will have the same >>>>opinion about it like me. >>> >>>Repetition of your case: >>> >>>>4. I also published my view, that computers should stick to normal >>>>theory >>>> >>> >>>> lot of nonsense lines that have nothing to do with >>>>opening theory and you'll need weeks to 'repair' this, maybe than you'll >>>>understand me >>> >>>*should*, *normal*, *need weeks*. >>> >>>IMO your argument boils down to: it tkes a *lot of time and effort* >>>fighting off book cracks, so we *should* be *moral* and use a *normal >>>book*. >>> >>>1. IMO it is impossible to exhort the mass of programmers to this >>>position, therefore one or more will break it (intentionally or >>>otherwise). >>> >>>2. The 'moralising' is out of place. Since all the opponent program is >>>doing is adding chess knowledge to his program - this cannot poissibly >>>be an immoral act. >>> >>>3. The definition 'normal book' is another 'morality' exertion. What is >>>normal ? >>> >>>The effect of this entire argument is to attack and humiliate Mchess and >>>supporters - that's actually the imorality. >>> >>> >>> >>>Chris Whittington >>> >>> >>> >>> >>> >>>>Regards, Jeroen >>>> >>>> >>>>>Only they are not cooked, they are auto-learnt. The difference is in the >>>>>intention. >>>>> >>>>>Further, why is it that the Rebel team are always so busy assaulting >>>>>Mchess ? First it was Ed several months ago with a major anti-Mchess >>>>>campaign; now you seem to be acting as proxy and doing the same thing. >>>>> >>>>>So he leaves the book with a plus score. So what. CSTal leaves book with >>>>>major plus scores, major minus scores, it happens; and I've never cooked >>>>>a book in my life. >>>>> >>>>>Plug the holes in your book. And thank Mchess for pointing them out to >>>>>you. >>>>> >>>>>Chris Whittington >>>>> >>>>> >>>>> >>>>> >>>>>> >>>>>>Regards, Jeroen >>>>>> >>>>>>> >>>>>>>I want to defend Mchess. >>>>>>> >>>>>>>Because, in these arguments over the past few days, I'm reminded of >>>>>>>Stalin's dictum from the 1930's. He said that *intentions* were >>>>>>>irrelevent. If the *result* of your actions were counter-revolutionary, >>>>>>>then you were a counter-revolutionary - and should therefore be shot. No >>>>>>>matter that you were trying to fulfill the plan if you made a mistake >>>>>>>and failed - you were to be shot. >>>>>>> >>>>>>>Now Mchess has a learning feature - it tries an opening; if it comes out >>>>>>>of the line with a minus, it remembers and tries another move later. if >>>>>>>it comes out plus, it remembers and extends the book. This way it builds >>>>>>>a book where 'bad' theory gets rejected; and a new Mchess idea gets >>>>>>>tried. If the 'new' idea works, it becomes part of the book, Hence the >>>>>>>later computer games of mchess where it plays as per some Gm or Im game >>>>>>>up to move 38, amd then there's another move, never seen before, or >>>>>>>other move series never seen before. So Mchess extends chess knowledge >>>>>>>via autoplayer games. They then release with the new book; and the ng's >>>>>>>start to skweak. >>>>>>> >>>>>>>The *effect* is counter-revolutionary, while the *intent* was greater >>>>>>>knowledge. >>>>>>> >>>>>>>You guys argue to shoot Sandro Nechi. Instead you should be applauding >>>>>>>him. >>>>>>> >>>>>>>Chris Whittington
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.