A reader asked a question about the following position from Polugayevsky - Nezhmetdinov, Russian Federation Championship, Sochi, 1958.
I suppose he was asking the question why, if 27.Ng1 results in a perpetual, doesn’t 27.a4 also lead to a draw. I believe he was making the incorrect assumption that Black would play the same sequence of moves against both moves.
First, after 27.Ng1 c5+ 28.dxc6ep Ned3+ 29.e5 Bxe5+ 30.Kc4 b5+ 31.Kxb5 Rb8+ 32.Ka4 Nxb2+ 33.Ka3 Nc4+ 34.Ka4 Nb2+ the game is a draw.
Second, after 27.a4 c5+ 28.dxc6 bxc6 29.Bd3 Nexd3+ 30.Kc4 d5+ 31.exd5 cxd5+ 32.Kb5 Rb8+ 33.Ka5 Nc6+ 34.Ka6 Ndb4 mate
But if White plays 27.Ng1 and Black tries the same attacking moves then 27.Ng1 c5+ 28.dxc6 bxc6 29.Bd3 Nexd3+ 30.Kc4 d5+ 31.exd5 cxd5+ 32.Kb5 Rb8+ 33.Ka4 White can’t play the K to this square if he plays 27.a4 and his K has to go to a5 and gets mated. I should note that in this position Black still has a winning advantage.
The fly in the ointment is that in this line is that White does not play 29.Bd3? but instead plays 29.Nxf3 and after 29…Nxf3+ 30.Kc4 Bxb2 31.Rb1 it is White who is winning.
Why did the authors not mention a better defense than 27.a4? There are several possible explanations. I don’t know when these books were published so can’t be sure how strong engines were at the time or even if the games were checked with engines.
The same game appears in The World’s Greatest Chess Games by Burgess, Nunn and Emms published in 1998 and they did check all the games using Fritz. Their analysis suggests several defenses, the first choice being 27.Ng1, but they commented that none of the defenses are sufficient. Of course that’s not true because after 27.Ng1 it’s a perpetual check.
I tried the position on Firebird, Fritz 10 and Rybka Human and they all quickly arrived at 27.Ng1 as being best. After 27.a4 they all found the mate very quickly. So it would appear the best defense was missed because of faulty human analysis or, if an engine was used, it wasn’t very strong by today’s standards. Or possibly it wasn't given enough time to ponder the position.
This is the reason why GM’s never completely trust engines and Mark Buckley said in Practical Chess Analysis, “always challenge the annotator.”