How a computer error helped Deep Blue beat humanity's best chess player

What I've always found so interesting about this story is how singular a moment this was, where humans and machines could play a meaningful game of chess. For millennia before, a chess-playing machine was nonsense. Today, chess computers so completely outclass human players that mobile phones can win high-level tournaments. There was a period of no more than five or ten years where the best computer could play the best human, and have the result be in doubt, and it happened right around the turn of the millennium.

I've always been of the opinion that the computers should have to play through qualifying tournaments to get to a "title" match like this.

IBM had thousands of matches to prep their player and Kasparov had none. This disparity is crazy for top level chess.

I've also never heard about this match being discussed in terms of the FIDE-PCA rift, which may have given the challenger more leverage as IBM could shop between two champions. With a unified chess world, Kasparov might have been able to put the screws to IBM for a more "fair" match.

Direct link to documentary http://espn.go.com/video/clip?id=espn:11694550

It's interesting, but it's not news. Nate Silver published this information in his book The Signal and the Noise back in 2012.

"After the loss Kasparov questioned whether Deep Blue's team had cheated in order to beat him, but in a fascinating new short documentary by FiveThirtyEight and ESPN, it's shown that the momentous victory — and the notorious 44th move that led to it — was actually the result of a computer error."

But the video actually says the error occurred in the first game, which Kasparov won, rather than in the second, which he lost and after which he suggested he had been cheated.

For a different spin on films about chess-playing computers, check out Computer Chess, a mumblecore mockumentary about a 1980s computer chess tournament:

http://www.imdb.com/title/tt2007360/

Can it really be called a "computer error" for a non-deterministic program?

It beat Garry Kasparov in multiple matches, it's probably better than most chess programs at the time.

More like.. The Ghost in the Machine.

This must be a mite frustrating.

why wrong? 1.)correlation does NOT imply causation 2.)psychology of FEAR mode can strike ANY chess player 3.)the computer at that time will lose because it does NOT understand the art of deception and 'feints.' 4.)IBM Deep Blue is moving to poker, which is a MUCH better test.

5.)observations are NOT double blind.

6.)highly useful IMHO (at least for me as a child) is to IMAGINE the opponent is 'genius' and then 'stupid.' Simple thought experiment. Even the dog with low IQ is able to predict the owner's moves.

7.)It's obvious that even basic 'stupidity' is mentioned on HN, the verge and even the Nate Silver book. Why?

8.)time or tempo is important and more so with high level champions. So, 'losing tempo' by 'computer error' or 'harmless rook move' is BAD.

9.)Obviously, it is far worse for a HUMAN PLAYER, since it means that the opponent has lost its train of thought. For Deep Blue it means BAD CODING.

note: based on similarities to my experience as child chess champion.

PS. the definitie book still NEEDS TO BE WRITTEN, including some open sourcing of code.

PPS. generally the only suggestion I can make to improving chess ability is YOGA or relaxation techniques. Why?

the key problem IMHO is stress cause by 'context switching' between frames depending upon 'models.' i.e. model of attack or 'strategic attack' or 'feint attack', etc.

the key question is: the computer does NOT switch between modes but has one global mode, IMHO. as a comparison to human player.