Two weeks ago a study by Oleg Pervakov was presented at The
Chess Caf. I consider him one of the world's best chess
composers. But five years ago I simply did not know his name. In
1994 Yohanan Afek (the Israeli master and composer, see my
posting on him at The Chess Cafe) showed me another Pervakov's
endgame study. I was so impressed that I decided to publish it in
my chess column in La Pense Russe. But then something
unexpected happened... But first things first... (See Diagram)

Diagram 
White: Kg2, Ra7; pawns - a2, a3, b7, d7, f2
Black: Ke6, Ne5; pawns - d5, e2, f5, h5, h2
White to move and win

It is not easy for White to win. How can he repel Black's threat
1...h1Q+ 2.Kxh1 e1Q+? Evidently, White first underpromotes... 

1.d8N+ Kf6 2.Ra6+ Kg5 3.Nf7+! Kh4/i 4.Ra4+ d4!/ii 5.Rxd4+
Ng4 6.Rxg4+ fxg4 7.b8B!/iii h1Q+ 8.Kxh1 e1Q+ 9.Kg2 Qe4+
10.Kh2 g3+!/iv 11.Bxg3+ Kg4 12.f3+!/v Kxf3 13.Ng5+ Ke3
14.Bf2+!/vi Kd3 15.Nxe4 Kxe4 16.a4 Kd5 17.a5 Kc6 18.a6 Kc7
19.Ba7!/vii Kc6 20.a4!/viii

i) After 3...Nxf7 4.Re6 Black's pawns are stopped. 
ii) Fantastic, but Black is playing for stalemate! 
iii) Forced since after 7.b8Q? h1Q+ 8.Kxh1 e1Q+ 9.Kg2 Qe4+
10.Kh2 Qf4+! 11.Qxf4- stalemate! 
iv) The only defence against the threat of Bg3 mate. 
v) Extremely elegant! 
vi) As will be soon seen, the necessary finesse.
vii) That's why White played his bishop on f2. 
viii) Mission accomplished! It is owing to this small pawn that
White wins the game. 

Just before I was about to publish this study, one of my Parisian
friends proposed to check it up with the computer program "Chess
Genius". In the position arising after 8.Kxh1 "Chess Genius" gave
the following variation: 8...Kh3! 9.Ng5+ Kh4 10.Bg3+ Kxg5
11.f4+ gxf3 12.a4 Kf5 13.a5 Ke4. So, it is a draw?! Frankly
speaking, I was quite upset. It is always with regret that you watch
a masterpiece crumble like this... So, Pervakov's gem was cooked?!
Deep down, I was about to agree that the computer was right, but it
so happened that some months later Afek came to Paris again and I
showed him what the computer had found. The Israeli master was
also upset at first, but a bit later, after some painstaking analysis,
he found "the refutation of the refutation"... (See Diagram)

Let us continue the computer's drawing line14.a6 Ke3 15.Kg1 Kd2
16.a7 e1Q+ 17.Bxe1+ Kxe1 18.a8Q f2+ 19.Kh2! f1Q 22.Qh1!!
The proof of Pervakov's pudding! White exchanges the queens and
again pawn "a" is the 'killer'!

Since I have seen many studies by Pervakov, but this one, the first I
saw, is my favorite. The length of the variations, no similarity
between the initial and final positions, underpromotions, economy
of material - all these make this study a true work of art.