<Over the last 8 or 9 years the Bayonet Attack has been one of White's main success stories in the King's Indian. Prior to this 9.b4 was generally thought to be a poor relation to the main alternatives in this position, 9.Ne1 and 9.Nd2. It was rarely seen in top class chess. Strange, you may think, as 9.b4 is the obvious way for White to force through c5-c5 as quickly as possible. In this position the pawn structure dictates that White will attack on the queenside. The reason why this move was not trusted was that it allowed Black to play the active 9...Nh5, whilst after moves like 9.Nd2 or 9.Ne1 the black knight, in order to get of of the way of the f-pawn, would have to retreat. Attitudes began to change after White discovered the move 10.Re1 (in reply to 9...Nh5). The simple idea is to meet 10...Nf4 with 11.Bf1. The bishop on f1 is very well placed defensively and it turn out, that despite his active appearance, that the knight on f4 is quite poorly placed. There is nothing for it to attack and it can even get in the way of blacks kingside play. For example, the traditional pawn storm with ...f5-f4 is not possible with the knight on f4 and black will also have to be constantly on the lookout for white playing Bxf4 at a favourable moment. It took a while, and cost an awful lot of points, before black players appreciated this. Meanwhile everyone was playing 9.b4. The old main lines, 9.Ne1 and 9.Nd2, just disappeared. Even Gary Kasparov got his fingers burned and the Bayonet Attack was the main reason for his recent loss of confidence in the King's Indian (he didn't play it once in his world title match with Kramnik, who happens to be one of the main advocates of 9.b4). Many other King's Indian specialists, such as myself, settled for giving up 7...Nc6. A whole new system with 7...Na6 was born almost solely due to White's successes in the Bayonet Attack.>

Bareev had played 16 Bxe4 against Radjabov at Enghien-les-Bains 2003 and Black had obtained a big advantage after 16..d5 17 Bd3..e4 though he later blundered and lost. The main idea behind 13 Rb1 is the exchange sacrifice 19 Rxe5!. 21 Bb2! was Shirov's improvement over 21 Bd4 and 21 Qd2 that he had discovered when working with Landa in 1997 and had finally got the chance to play here. 32 Be4+..Rf5 33 Qxe7..Nxe7 34 g4..Kf6 35 gxf..b5 would only have been good enough for equality.

Shirov after 35 Qg4+
"After this important check there will be no fortresses, so White can begin converting his advantage into victory."

36..a6?! created a new weakness; 36..Ke8 would have been better as taking the a-pawn allows an easy equality after 37..Re1+.