Squares in the King's Indian

     While browsing the November 2011 Chess Life there was an interesting article on the US Open in which seven GMs tied for first: Nakamura, Gelashbili, Gareyev, Ramirez, Kacheishvili, Lenderman and Zapata. 
     The article featured some games with superb annotations by GM Alejandro Ramirez. What caught my attention was the last round game between Ramirez (2674) and GM Julio Sadorra (2561) in which Ramirez discussed a K-Indian position and the importance of squares and black's backward d-Pawn which is frequently seen in the K-Indian. 
     In his classic Zurich 1953 David Bronstein revealed the secret of black's backward d6-Pawn that often appears on the semi-open d-file. It should be weak, but it often proves a tough nut to crack and if white injudiciously captures it bad things can happen. 
     Very often white will have a N on d4 and it seems a simple matter to simply retreat it and take black's P, but as Bronstein pointed out, d4 is precisely where white needs the N to keep an eye on the squares b5, c6, e6 and f5. Sitting on d4 white's N also neutralizes the power of black's B on g7. The d6-Pawn can be captured, but only after white has taken precautions against all of black's possible attacks: ...a7-a5-a4-a3, ...Bc8-e6, ...f7-f5. Only then can the N be removed from d4, but in the meantime black has time to regroup. If the opportunity arises early in the game for white to capture the P on d6 a good rule of thumb is, "Don't." 

     Also, sometimes when the center is closed (white Ps on c4, d5 and e4 with black Ps on d6 and e5) you'll see black play ...Nd4. This often gives white the opportunity to capture on d4 and after black replies ...Pxd4 white may even be able to win the P. Often though white refuses to capture on d4. The reason is that black benefits substantially from his e-pawn's disappearance. The half-open file allows him to exert pressure against white's pawn, forcing some of white's pieces to awkward squares, and the lever ...f5 often threatens to demolish white's center. The open a1/h8 diagonal means that, assuming black can regain his P, his dark square bishop will be no liability in an ending. The recapture ...e5xd4 will also make the e5 square available to black's pieces. It's not all in black's favor though; white gets the d4 square. 
     What got me curious in this games was how the GMs opinions matched the evaluations of Stockfish 8 and Komodo 10, especially since engines, even today, do not often give correct evaluations of the K-Indian.
     When analyzing the below game, Ramirez gave very little analysis and mostly just gave his general impression of what would have happened if he had played a different move. The position after black's 19th move was especially interesting because the GM thinks black's position was "just awful", but the engines disagree, evaluating the position as almost dead equal. As I pointed out in my notes to the game, this is what makes modern engine-assisted correspondence chess such a difficult challenge...discovering which, if any, of several moves of nearly equal value might actually be better and is the engine evaluation always correct or is there something in the position that's over its horizon? Just to get some idea of the general course the game might take I ran some Shootouts from the position after 20.Nd5 at 17-25 plies using Stockfish 8 and Komodo 10.  Both engines produced identical scores of +2 -0 =3, so I think it's safe to say that the GMs opinion of the position is closer to the truth than the engine's equal evaluation.