DrMAL: In Nakamura vs Van Wely, 2007 black tried 11.Na5 to prevent 12.Nf5 and also allow option to take B on b3 (which he probably should have done, see comments there). Here, Salov went back to original position in Fischer vs Larsen, 1970 and decided to decline sac, good option unless one knows exact move order one should not accept sac very sharp position required extreme accuracy.
Houdini_20_x64: 29/74 1:43:21 57,763,591,478
0.00 12. ... exf5 13.Nd5 Qd8 14.gxf5 Kh8 15.Rhg1
-0.27 12. ... Nc5 13.Nxe7+ Qxe7 14.h4 b5 15.f3
Engine score is much less useful than lines especially move order. Some moves such as 14...Kh8 versus 14...a5 (if sac was accepted with 12...exf5) make little difference but most are quite critical. 13...Nxe7?! instead of 13...Qxe7 was more on the critical side, it does not lose immediately but gives white significantly more initiative. Anand took maximum advantage with 14.Qd2! to play on dark squares (14.f3 for different plan was also strong) here 14...Nxb3+ to eliminate B was maybe best but 14...Rd8 could not protect d6 but in lining up R with Q could pin B against Q (particularly with 15.Bf4 chosen to focus on center starting with taking pawn. Alternative 15.h4 for K-side attack was also very strong). Now, after 17.f4! black passed up another opportunity to eliminate B on b3, counterattack with 17...b5?! was mistake, too late for that, black needed to focus on defense.
18.f5! was certainly strongest here with this, 18...Ne5 19.Bd5! forces 19...Ned3+ with 20.cxd3 exd5 21.Be7 Rd7 22.exc5 Rxd5 23.Qg2 Rd6 24.Qxc6 Rxc6 25.Kb1 for big advantage. It is long line, I know, but with few options player like Anand could quite easily see. He chose simpler positional route with 18.e5 I guess it was simply more comfortable and certainly cannot be criticized. With Salov's insistence on advancing pawns after 20.Nd4 forcing 20...Nxb3+ Anand got similar advantage anyway. 22.Rhf1! was similarly much stronger but Anand chose 22.f5 and position simplified to much very small edge, probably draw.
But Salov, who played quite accurately up to this point, finally made big error with 30.Bf5? losing pawn. Endgame from this was matter of good technique, and Anand is certainly not lacking there! In case there was doubt 27...Bb7? (instead of Ke5 or Kf5) removed it, with 38.Ne4! black could resign.