DrMAL: In this game Short used Charousek variation (3...Be7) that has several subvariations. 5.Bf4 started Alatortsev (5.Bf4 c6 6.e3 Bf5) which often prompts 7.g4! Be6 (or 7...Bg6 not as good) for strong and sharp attack as in recent Nakamura vs Aronian, 2011 (ignore silly criticism of 7.g4! there), this is main reason why 6...Bd6 often gets better results. Kasparov chose 6.Qc2 which can transpose to Alatortsev as it did, this was interesting alternative then. 6...Bd6 is most popular but results have been unimpressive and opening is still not played much, maybe 6...Nf6 is more promising. 6...g6 was interesting then but now considered quite poor and not really played at high level any more. Spassky originally tried it and lost Korchnoi vs Spassky, 1968 but maybe best example is famous Karpov loss Topalov vs Karpov, 1998.
9...c5 attempt at sharpness makes matters for black worse (10.Bh6! is obvious reason), best was simply to castle. Game is not good example of opening play, miniature is included because it shows excellent example of Kasparov attacking technique. On move 10 white has other options here is deep computer eval.
Houdini_20_x64: 32/78 31:03:42 1,177,070,553,482
+0.44 10.Bh6 Nc6 11.Bb5 cxd4 12.exd4 Bf8 13.Nge2
+0.36 10.Bb5+ Bd7 11.Bh6 Bxb5 12.Nxb5 cxd4 13.Ne2
+0.29 10.g4 Be6 11.Bh6 cxd4 12.exd4 a6 13.Bd3
+0.23 10.dxc5 0-0 11.Nge2 Bxc5 12.Nd4 Bd7 13.Bg5
+0.23 10.Nge2 0-0 11.dxc5 Bxc5 12.Nd4 Bd7 13.Bg5
Position here was worth deep computation, I was interested in how move order affects lines, such as idea of 10.g4 intermediate move since it had not been played earlier. It also keeps with old saying: see good move look for better one. Bb5+ was also interesting idea on move 10 or 11 but Kasparov wanted B on g2, allowing 11...a6 to prevent it. His technique here can be summarized as "sharp and most direct" avoiding anything fancy.
Computer offers 12.Qe3 as best to pin B and centralize Q for pressure on e-file where 12...Be6 is also likely response. However, this is less sharp than 12.g4! so it is no surprise what Kasparov chose. Computer also gives long castle as transposition (12.g4 Be3 13.0-0-0 Nc6) but Kasparov was more interested in sharpness and tempo. Likewise, intermediate move 12.h4 Nc6 13.g4 was avoided, consistent with "sharp and most direct."
This immediately paid off, after 13.Nge2 to get N into attack Short chose 13...Nbd7?! instead of 13...Nc6 perhaps this was aimed at Bf8 with N to recapture then move away for castle. 13...Bf8 was also good, after B swap black K can go to g7. It seems Kasparov was almost clairvoyant about not playing Bb5 earlier 14.Bg2! was now excellent position for B. 14.Nf4! is also clearly strong but bishop on e6 is not going anywhere and Kasparov practices another key principle in his technique "don't rush!" 14...Nb6 clarified intention of 13...Nd7 to maneuver Nc4 but this requires three tempos, and worse 15.b3! puts N in bad position on b6 anyway, with no future and away from battle.
This shows another key principle to Kasparov technique "get opponent pieces away from battle" that often appears as some sort of magic in his games. After 15...Rc8 (strong) 16.0-0 Rc6 to try and create some sort of counterplay 17.h3! reinforcing g4 underlines black complete inability to do so (computer gives 17.a4! showing white dominance all over board). Here 17...Bf8 seems clearly best (earlier too) but again Short chooses slower 17...Nfd7 for some sort of maneuver. Kasparov took advantage with more important manuever Nd1-Nf2 readying N for attack.
Short again avoided 18...Bf8 I guess his otherwise pointless 18...Rg8 was aimed at going g5 as he shortly did. But 19...f5?! was again rather dubious, part of plan for K-side counterattack. Black pieces are not poised for this and 20.Rae1! to ready last piece, being consistent and ignoring was strong reply. Now 20...g5?! was probably decisive but 21...Bf7? instead of 21...Bxf5 insured white victory after 22.Ng4! black is in virtual zugzwang here. With 22...Bh5? desperation move 23.Ng3! made hopelessness even more obvious, resignation was clearly best.