From all the bad press on 8...b4 it's unexpected that top engines evaluate it as sound, especially if followed by 9...Bb7 instead of immediate 9...Nxe4 a slight difference. For example, even at d=22, Rybka evaluates 8...Bb7 at -0.19 and 8...b4 at 0.07 as second best. Moreover, the game demonstrates its soundness. Fischer plays exactly the same choice as Rybka 4.1 under deep computation move by move, until 16.Bxc8 where Rybka's first choice is basically a toss-up.

17.Bf4 is the first real departure where 17.Rf2 and 17.c3 are evaluated as nearly the same and 17.g3 a very close third. Here, Rybka's evaluation drops by 0.6 to +0.3. These computer results have meaning to me because there are too many continuations to evaluate confidently. Both players are simply amazing to me here. I don't think 17.Bf4 created a missed opportunity as Fischer later stated, white's attack is far from decisive.

After 20.Bxb8 the position is evaluated dead equal where 20.Bxe7 adds around 0.2 in white's favor. After re-capture, Fischer could have played 22.Rae1 for a draw but instead he played the only other good move, 22.Qc6+ giving about half a pawn worth of advantage to black. 24.Rxf7 instead of 24.Qf6 seems to be the first real mistake up to now, evaluated as giving black a whole point more.

<Few more details from that game , Fischer vs Tal, 1959 . Move 22 , and fischer is thinking over , probably winning , 22. Rae1 ! . Lets give word to Mikhail Tal :

<"Every player has his own habit: one will first make his move and then write it down, while another will do things the other way around. Incidentally, in recent years Fischer has actively objected to this 'other way round', expressing the opinion that a scoresheet is not a black-board for writing down exercises. However, in our game Fischer first wrote down the move 22.Rae1!, without a doubt the strongest, and wrote it not in his usual English notation but in European, almost Russian! Then he not very deftly pushed the scoresheet towards me. 'He's asking for an endorsement', I thought to myself, but how was I to react? To frown was impossible, if I smiled he would suspect 'trickery', so I did the natural thing. I got up and began to calmly walk up and down the stage. I met Petrosian, made some joke to him, and he replied. The 15-year-old Fischer, who was essentially still only a large child, sat with a confused expression on his face, looking first at the front row of spectators where his second was sitting, and then at me.

On move 22, Rybka 4.1 evaluates (Rae1 can get perpetual):

[+0.00] d=18 22.Rae1 Kd8 23.Rxe7 Bxe7 (0:11:04) 107755kN

[-0.61] d=18 22.Qc6+ Rd7 23.Rae1+ Be7 (0:14:26) 142785kN

Whereas on move 24, the evaluation is (lines also truncated):

[-0.61] d=18 24.Qf6 Kd8 25.Qxf7 Re8 (0:21:34) 177060kN

[-1.53] d=18 24.Rxf7 Kxf7 25.Qe6+ Kf8 (0:22:53) 189544kN

Reading in My 60 MG's it does seem he regrets not taking more positional care with his weaknesses on g2, a4, the diagonal a8-h1 and a probable tempo loss with the K on g1. He specifically mentions moves like 15.Bf3 and 17.c3.

Fischer ego was colossal and his analysis in book tends to feed into this, I guess he never imagined computers being able to easily expose motivation behind his conjectures. 18.Qf3 offering Q trade or sac was good idea to try for something but Tal's play during next few moves made it backfire, pressure was back on Fischer to not lose from it.

And Fischer immediately started to go wrong with 22.Qc6+ instead of positional move 22.Rae1 but he corrected himself next move, showing he was mortal like the rest of us. Attack gave nothing but small advantage to Tal who had played perfectly. When this became apparent, 24.Rxf7? was reckless nosedive against soon to be WC Tal, nice try kid but come back next year LOL.