|Amarande: <JonathanJ: <newton296> i don't think that Ng5 is an easy win against a 2200+ player if you aren't even able to find this queen sac :P>|
One has to consider here whether newton even bothered to *look* for the Queen sac.
If one first analyses 22 Ng5, one will of course see the threat of 23 Qxf7+ clearly. One then checks to see if Black has a defense against this, and sees quickly:
* if Black tries moving the King away, 22 ... Kh8, 23 Nxf7+ mates in four, by the classical smothered mate pattern.
* if Black tries simply taking the Bishop, it is enough to notice that 23 Qxf7+ wins the e8-Rook for it while in no way abating White's attack, to see that this is hopelessly insufficient.
* that the Black Rook that is involved in the threat posed by 23 Qxf7+ is shy a safe escape.
* thus, Black's only hope is to try to protect f7 further. If he tries to do this with the Rook, 22 ... Rf8, then simply 23 Bxf8 Nxf8 24 Qxf7+ Kh8 25 Rxc8 and Black can resign with a clear conscience.
Thus he can only try 22 ... Ne5, which is refuted most beautifully. 23 Bxe8!! is the move, completely ignoring the pin that Black relied upon to save his Rook! 23 ... Rxd5 now fails to 24 Rxc8!!
click for larger view
The immediate threat is 25 Bxf7 mate. However, because this is a double check there is precious little Black can do about it: he is already guarding f7, and can attack c8 by offering his Queen, but of course both of these are futile in the face of a double check threat. f7 and h7 are guarded by the White Knight, so Black can only hope to escape by moving his Bishop, to give his King room.
Four defensive tries are possible:
a) 24 ... Bh6 blocks a critical secondary escape square and White mates by 24 Bxf7+ Kg7 25 Rg8#.
b) 24 ... Bf6 is interesting, as it leads to a mate involving a further echo of the same theme: 25 Bxf7+ Kg7 26 Bf8+ Kh8 and now not 27 Bh6+? Bd8, but 27 Ne6!!, and again White threatens to mate by a double check with a Bishop (28 Bg7#). This time, though, Black will not survive - the only way to give the King room is to move the h-pawn, but after either 27 ... h6 or h5 there is 28 Bg7+ Kh7 29 Rh8 mate.
c) 24 ... Bh8 25 Bxf7+ Kg7 26 Bxd5 and Black can resign; he has lost far too much material for the Queen he has won, and White still has both a strong attack and the passed pawn d6 besides.
d) 24 ... Bf8 25 Bxf7+ Kg7 26 Bxf8+ Kf6 (if 26 ... Kh8 27 Ne6 leads to the same mate as in variation b above) 27 Bxd5 and again Black is too far behind on material to live.
All this means that after 23 Bxe8, Black cannot take White's Queen and must instead do something about his attacked Bishop to prevent the deadly Rook incursion. Naturally 23 ... Ra8 is tantamount to resignation, as Black has simply lost a Rook and given up his only counter threat, while White's attack remains strong as ever; therefore, only 23 ... Bd7 is to be considered. 24 Bxd7 is the reply and Black may as well take the Queen now, else he has given up a Rook for nothing and he has no more tricks. 24 ... Rxd5 25 Rc8+ Bf8 26 Rxf8+ Kg7 27 Bxa4, and Black is out of the woods for now, but White has a Rook and two minor pieces for the Queen, still the dangerous passed pawn on d6 (Black has a passed pawn too, but at b7 it's very little of a threat) and Black's King is by no means safe, so Black is bound to lose in due course.
Therefore, analysis of 22 Ng5 is enough to establish it as more than sufficient to win, and that there are no surprise counters to it. At this point one normally plays such a move. As Chernev once said about Chekhover vs I Rudakovsky, 1945, "leave the brilliancies to Alekhine and Keres." The idea being, of course, that once you see a clear way to win, it's generally better to simply execute it, than to go off looking for brilliant moves (even if they are correct, which is the case in both this game and the position Chernev's famous comment referred to). Why take the risk the brilliancy might be flawed, and even if it turns out not to be, why waste the extra time to look for it?
FWIW, I did see 22 Qxf7+, and thus, similarly, didn't even consider 22 Ng5 until the topic came up here on the board :)