How did Gasparov 1-1-1 Deep Blue?

I read that Deep Blue, the supercomputer, beat Kasparov once, drew once, and then actually lost one to Kasparov. K. asked for a rematch but Deepblue's IBM makers refused, claiming that Deep Blue had more important tasks ahead.  How did Kasparov beat Deep blue? How could a supercomputer with so many trillions of possible chess situations lose to a human being. Can anyone give me a better reason that the magic of the human soul?

Well, I don't know the process behind which deep blue examines and selects moves, but assuming that it did it through brute force, it means the engine was examining many moves that are easily dismissable possibilities.  In this way much of the effort the engine is doing is inefficient.  So these weren't all "good" moves that Deep Blue was examining.  Most were very poor moves that any average level player could dismiss.

The GM's mind on the other hand has learned the patterns that dictate a good move.  This way the GM's mind is not relying on going through all possible moves, but using their knowledge of patterns to see the relevant aspects of the position and the patterns that dictate what a good next move will be.  So these super calculating ability is not required because the problem is examined differently. Incredible amounts of unconcious calculation are still being performed, though.

So I wouldn't say its the human soul, but that the human mind is an incredible machine that used a different information processing method.

Also, computers are programmed by humans.

Cultivate we are talking trillions of positions. The magic of the human soul? Well it probably had something to do with the soul.

Yeah, you're probably right.  Kasparov was one pretty powerful chess soulja...

"

Also, computers are programmed by humans."

I never got that argument. That's like saying smoking a plant is healthy because it's natural. I'm pretty sure computers already surpassed humans on many levels, though not all of them, even though we made them.

Well, I see the argument more as suggesting that one cannot create something that surpasses its maker.  Which obviously isn't true. 

I think people are saying how can someone teach an engine chess principles and strategies that the designer does not understand or know in the first place.  I do not believe that engines are programmed in such a way as to explicitly program principles this way.

cylons dudes

A saying in my Kasparov vs Deep Blue book would go nicely here -

"You cannot trust the output if you cannot trust the input." (input referring to the programmers and GM move suggestions which built up deep blue's chess ability. As the input is only done by human hands, it is therefore falliable)

Basically, computers don't evaluate positional advantages as well as humans do.  Humans, knowing this, play positionally.

It's pretty much impossible to out-tactic a computer, and they have endgames solved by a "this position is mate, look one move back, then one more" formula.

I play a lots of otb games. and then I analyse them on my chessmaster 9000. I played a lots of moves, for which he say that aren't in top 5 moves, but when I put them in a game, he "changes his mind" and I stand much better in a game than he said for the best move. that could be the reason. I saw a position from 1964 I think. It was an endgame with same number of material. Black won effectively and computer foun that out after 4th move. laim =P

Kasparov have been optimistic before the game started. He used non-theoretical openings in order to avoid early confrontation, to positional advantages and then he was confident that his calculation would stay at the high level once the confrontation occured.

There was another topic on this on monday...

http://www.chess.com/forum/view/game-analysis/deep-blue-beat-garry-kasparov-in-1997

The only possible explanaton is that Kasparov is himself a robot.

I've read a book about the match and other people's comments. It seems that computers are particularly strong in certain types of position where calculation is important i.e "open" positions where tactics abound. In "closed" positions, where the pawn chains are locked and there is no immediate threat to either side, computers can drift into poor positions without realising it. I have also heard that the computers' Opening Book can be a weakness as it relies upon the input of moves by humans. If the human puts in an inferior move, the computer will slavishly follow it's Opening Book and gain a poor position. On the human side, Garry is probably the best player of all time, and his instinct honed by knowledge of similar positions he has played allowed him to hold his own against Deep Blue. With the advent of even stronger programs e.g. Rybka Shredder and Deep Fritz, the moment has now arrived when computers will beat any human opponent they encounter almost every game. It's a far cry from the mid 1980's when my Fairchild chess microcomputer did not understand how to mate me even when left with a Queen and a King against my King!