Ulhumbrus: 5...Be7 conserves the bishop pair.
14 Bc3? looks like a mistake. After 16 bxc3 Black has an isolated d5 pawn, but White's Q side is broken and Black has an advantage in development. Which suggests the question: how does Karpov manage to turn the tables? 18 Kd2 provides the beginning of an answer: White's K is several moves ahead of Black's King in development in the ending. White is playing in fact with an extra King. Which suggests the question : Can Black make the White King irrelevant?
19..Be6? is inconsistent with ...Nd8 in one respect: it blocks the use of the square e6 for the manoeuvre ...Nd8-e6 followed by ...Nxc5.19...Bd7 keeps e6 free for the N. With 21 a4 Karpov has minimised the effect of his broken Q side and his K helps to defend c3.
23...a6 weakens the point b6. What is less obvious is that it makes d5 vulnerable to attack, for a White R can now use the point b6 to get to d6 and attack the d5 pawn. Instead of 24...Re5, 24...Re7-c7 makes it possible to play the Black King to e7 without taking the risk of trapping the Rook on the e file. The result of Black's inability to play ...Ke7 is that the point d6 is left undefended and so White manages to play Rd6.
One might say that it is because of the move a5 that the move 24... Re5 is a mistake,although this is hardly obvious at first sight. The reason for this is that a5 prepares the invasion Rb4-b6-d6 by inducing ..a6 while ..Re5 precludes the defence to the threat of Rd6, namely the move ...Ke7. Instructive, is it not? Perhaps we shall learn a few more things from Karpov's wins in this tournament.
Following 30 Rd6, the move 31....Ne6? is too optimistic, offering White the d5 pawn, after which the complications which follow do not in fact win the material back; After 39 g5 White has an extra pawn and Karpov's technique will suffice to gain a win from it.