Game 2 of the 1997 Kasparov - Deep Blue match was the subject of much controversy and debate. Kasparov resigned in what looked to be a hopeless position. Shortly after, gossip began spreading like wildfire that Kasparov missed a draw by perpetual! News eventually got back to Kasparov about this and he was clearly distressed over having missed it. Indeed it may have set the tone for him later choking in the final game to lose the match. There would even be documentary footage covering the infamous missed draw by Kasparov. What is this all based on? Let's have a look at the game:
From the final position in the above game diagram, Kasparov resigned feeling he had nothing left to save the game. Shortly after, Frederic Friedel and others announce there was a way for Kasparov to force a draw by perpetual. News got back to Kasparov about this and he was devastated. Instead of resigning, the claim was Kasparov could have played 45...Qe3 46. Qxd6 Re8! 47. h4 h5! (or Qxe4!!) and now black can create a means of perpetual check no matter what white tries. Here's an example below starting from 45...Qe3:
So where was everyone going wrong back then? Well as I noted in the diagram above, 47. h4 is purely a bad move that indeed allows black such a means to create the draw by perpetual. The problem is computer engines were so weak back then that it seemed clearly the best move to make. Once again, people were far too trusting of the engine's opinion. Today, my quad-core PC with Deep Rybka 3 immediately discovers 47. h4 throws the game away to the perpetual draw. Instead, white needs to play 47. Qc7+ (or Qd7+) Re7 48. Qc6! whereupon 48...Qxe4 is met by 49. d6! hitting both the queen and the rook at the same time! Sequence below:
Wait, so does this means Deep Blue had the win afterall? Well, not exactly. What it means is there was truthfully no way for black to force a perpetual check draw. Rybka spots this resource in the matter of one second on my computer, but whether or not Deep Blue would have spotted it remains a mystery. The continuation after the Qc6-d6 sequence becomes a very difficult uphill struggle for black to try and reach a drawn endgame. Lets take a look at the entire game with the new theoretical endgame sequence in place to see the kind of troubles black must overcome:
In the final position, the game has reached a seven-piece ending. I do not have 7-piece tablebases, but I think everyone can agree we'd all prefer to play white from here ;-) Now of course, this entire endgame sequence is just a quick blitz-out of analysis and nothing entirely deep, but it's one example of black fighting to just barely hang in there. One wonders whether Kasparov might have played well enough in this endgame to hold onto the draw...