The position in the diagram, from Fischer - Tal, Bled 1961 (see that post for punctuation by Fischer and Kasparov and for a link to the game on Chessgames.com), is the prelude to one of the best known opening blunders in chess history. After the moves 1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 e6 5.Nc3 Qc7 6.g3, Tal played 6...Nf6. Fischer remarked,
Probably the losing move! Tal looked worried immediately having made it, but I'm not sure he was convinced he had really been careless. Correct is 6...a6.
and assigned the move a '?'. Kasparov gave it a '?!'. The difference in opinion between the 11th and 13th World Champions is significant and can be summed up by the question, 'At what point did Tal play the losing move?'
After 6.g2-g3
After 6...Nf6, the game continued 7.Ndb5 Qb8 8.Bf4 Ne5. Here Fischer commented
Tal took a long time on this risky reply. The alternative 8...e5 9.Bg5 a6 10.Bxf6 axb5 (not 10...gxf6 11.Na3 b5 12.Nd5) 11.Bg5 gives a clear advantage.
Kasparov gave the move 8...Ne5 a '?' and remarked,
Huebner did not agree with this: 'Fischer often lacks rigourness when evaluating two options in a cheerless position; to my mind, this is the case here. The text move loses by force; after 8...e5 9.Bg5 a6 10.Bxf6 axb5 11.Bg5, Black would still have had chances to offer resistance if he continues 11...Bb4 12.Bxb5 Bxc3+ 13.bxc3 d6'. [...] The weakness of White's Queenside greatly hinders the conversion of his extra Pawn.
Moreover, in the variation 10...gxf6 11.Na3 [instead of 11...b5], Black should go in for 11...Bxa3 12.bxa3 Ne7 13.Nd5 Nxd5 14.Qxd5 b5 with a somewhat inferior, but by no means lost position.
There followed 9.Be2 Bc5. Here Fischer added,
In the tournament book, Tal suggested the rather startling 9...Ng8, to avoid material loss. After 10.Qd4 f6 11.O-O-O a6 12.Nd6+ Bxd6 13.Qxd6 Qxd6 14.Rxd6 leads to a promising endgame.
Kasparov gave 9...Bc5 another '?', and stated that 9...Ng8, was 'Black's last chance', agreeing that 'he would at least have retained material equality'.
***
The preceding summary offers three candidates for 'the losing move': 6...Nf6, 8...Ne5, and 9...Bc5. Tal must have had an extremely bad day to make three mistakes in four moves. In the case of 6...Nf6, Kasparov wrote that
Tal intended the usual 6...a6 7.Bg2 Nf6, and had already written down 6..a6 on his scoresheet, but, by his own admission, 'roughly once a year it would happen that I would write down the first move of a variation, but make the second.
Nowadays the act of recording the move before playing it is against the rules, due to the rise of pocket electronic scorekeepers that display the resulting position. In 'Russian Silhouettes', Genna Sosonko also comments on this habit of Tal.
He always used to write his move before executing it on the board. [...] If he did not like the move, he would cross it out and write a new one. In his later years he used to say increasingly often, 'I even wrote the winning move on my scoresheet, but crossed it out at the last moment...' (p.28)
As for 9...Bc5, Fischer noted that the alternative 9...Ng8 allows a 'promising endgame' for White. After 14.Rxd6, the last move in the variation given by both Fischer and Kasparov, White is ready to activate his last piece with Rhd1, doubling the Rooks on the d-file. Black still has five pieces on the back rank and still needs to move a Pawn to develop the light squared Bishop. Given this significant lead in development plus the two Bishops, White's 'promising endgame' offers strong odds to achieve a win.
I checked the historical game databases and discovered that Tal's 6...Nf6, has been played many times. In the May 2003 issue of Chess Life, GM Andy Soltis commented,
Tal gave himself two question marks for 6...Nf6??. [...] Yet it has been played dozens of times since 1961, including by Judit Polgar -- and Igor Ivanov did it twice in one year. And there are 24 examples in my database when masters failed to find 7.Ndb5!.
I suspect that one reason these blunders keep coming back is that masters are always saying that you shouldn't waste your time on traps. You should be wasting your time on loftier stuff, like theoretical novelties on the 32nd move. (p.14)
Even after the correct sequence 7.Ndb5 Qb8 8.Bf4, I found four games that followed Kasparov's 8...e5 9.Bg5 a6 10.Bxf6 gxf6 11.Na3 Bxa3. These games ended in Black's favor by a margin of +1-3=0. While this is too small a sample to draw conclusions, maybe 6...Nf6 isn't a blunder at all. Instead of giving it '??', '?', or '?!', perhaps we should give it '!?'.