i have been told that the high end computer software these days can beat all but the best GMs and even then they can give them a run for their money - eg kranick and kasperov both drew their recent matches against computers.
so does this suggest that the games accummulated through the decades are 'flawed'? after all, the players were OTBing under pressured conditions. for instance, anderssen's evergreen game was found to be 'not correct' by lasker i think it was. what percentage of the GM games would prove to be flawed if subjected to computer analysis?
the second question though is whether the computer analysis is really 'correct' in itself. computers may find the 'better' moves, but surely that is true only within the framework their calculations can achieve. so even though computer analysis may find a GM's game as 'flawed', perhaps it is the computer analysis that is flawed, because it was unable to see the depth and rationale of the GM's ideas.
any thoughts on the above?
in friendship,
prad
One of the good points about Kasparov's recent volumes 'My Great Predecessors' is the re-evaluation of famous old games using fritz. Lots of holes are found etc.
However, I don't think this takes away from the beauty or value of these games. They have to been seen in the context of the time they were played. If humans haven't found the flaws in them for decades but a tiny chip can, then that says more about our technologiacal progress than our chess progress.
Originally posted by Phlabibitthey already do.
I wonder what would happen if programs like fritz and the such removed all their chess book info... played tons of games... and created it's own books from there...
Would they be all that different? I think someone should get on that right now.
P-
some use banks of all the old GM games as books,
any competitive computer these days learns from its own games as well.
for a really fascinating look at this stuff check out jon baxter's input into a program called knightcap - the program decided on its choice of how to apply midgame strategies. most turned up as you would expect. it loved knights on KB6! unfortunately the features had to be given to it, then it would just prioritise them. it used to play 24hrs a day on fics for a few months.
keep an eye on ssdf to see most of the best programs battle it out.
p.s. if you find an opening mistake by a computer - do not force the win imediately, fiddle around a bit, waste several moves, then it will not connect it's loss to the opening mistake. if it does connect it then it will not do it again.
Originally posted by flexmoreHey Flex,
they already do.
some use banks of all the old GM games as books,
any competitive computer these days learns from its own games as well.
for a really fascinating look at this stuff check out jon baxter's input into a program called knightcap - the program decided on its choice of how to apply midgame strategies. most turned up as you would expect. ...[text shortened]... onnect it's loss to the opening mistake. if it does connect it then it will not do it again.
Computers, you say.... Who doesn't like Kt-KB6!? 😉
I'm not sure I understood your post about chess computers.. I know they use books, but what if they were to NOT use the books and create their OWN books from scratch? No human hints from the start at all.
Would these "computer generated books" be useless against a modern GM? Can Computers "learn" chess better than Humans did??
I know computers improve their existing books as they go, but what about a book Computers made from scratch? I actually think it is kind of "crap" that these computers are using info created by humans to beat us! 😠... 😉
P-
Originally posted by Phlabibitthey have been done from scratch as well, no use of human input at all.
Hey Flex,
Computers, you say.... Who doesn't like Kt-KB6!? 😉
I'm not sure I understood your post about chess computers.. I know they use books, but what if they were to NOT use the books and create their OWN books from scratch? No human hints from the start at all.
Would these "computer generated books" be useless against a modern GM? C ...[text shortened]... nd of "crap" that these computers are using info created by humans to beat us! 😠... 😉
P-
GM level computers have played a massive number of games against each other. these games form the vast majority of the database for the computers to use to construct their opening book.
adding the human gm games really makes very little difference - there are just so few of them (tens of thousands is nothing in computer language).
these days i think the openings are well and truly owned by the computers, the humans get advice from them!!! this will soon become even more noticeable as the number and quality of games in their database grows exponentially (with computer speed).
a similar thing has happened in the endgame, computers now play many positions perfectly just by memory of there previous calculations. and of course calculate well in many situations. then again there is still more work to be done in the area of the middle game .... they almost seem human in their stupidity there sometimes.
as for Knight-KB6 that was just a demonstration of a feature of the board that the computer "discovered". i guess the enemy king is as good as dead if you can get your horsey there. other features it liked were of course pawns side by side, doubling the enemy pawns, castling and high mobility of peices.
I would love to see one of these "All Computer" opening books.... I can't think of anything they would do other than play openings like Queen and King Pawn Openings... Did they find the Bird, English, Sicilian Defense???
Hmmm?
Here is a game if you have not seen CM9000 vs. Fritz7. Check it out.
Game 318397
P-
i seem to remember knightcap liking centrecounter. although it experimented with some others. it found many standard lines. but it was not a real strong program. only about 2100.
perhaps looking at kasparov - kramnik games you are often looking at computer discovered openings anyway? i believe they train their openings against computers in a very serious manor.
i agree it would be interesting to see a major computer generated shift in opening technique, unfortunately a patzer like myself has not. perhaps stronger players recognise one.