Analyzing Magnus Carlsen's Endgame Technique (Part III)

Welcome ambitious chess enthusiasts! I'm glad you've come to Part III in my Magnus series in order to improve your chess strategy and practical endgame play.

There's nothing more precious then a game collection on the modern genius of the game.

As a reminder, I'm reading Hungarian IM Tibor Karolyi's ground-breaking work: Endgame Virtuoso Magnus Carlsen: His Extraordinary Skills Uncovered and Explained on the interactive Forward Chess platform. For further context on the book and Forward Chess, please see Part I. In the previous blogs I demonstrated the methodical book structure and provided a thorough introduction to Karolyi's work.

Today our game of the day is from Chapter 3: World Class Player. 

However, I selected a special example from the 2009 section of his games. A blitz game where he managed to draw an unpleasant endgame against Boris Gelfand...

According to Stockfish 10, (which I conveniently used on the interactive Forward Chess interface) Carlsen was losing but through human error in time pressure and cunning from Carlsen's part, he deservedly ended the game unscathed. 

Karolyi picks up his analysis starting from move 59 but for the readers I provided the full game. (It's not that important to see the opening from a blitz game)

Lessons:

• Carlsen understood what his final goal needed to be in the position in order to draw: He needs to exchange all of black's pawns off. Throughout the game Magnus played with this in mind. (59.f5!)
• Gelfand should've limited his opponents counterplay by converting his decisive advantage into a trivially winning knight vs. knight ending, (by sacking one of his knights) 2 healthy pawns up, without any trouble.
• Magnus played purposefully with his king and was quickly able to run to the queenside in order to neutralize black's dangerous outside passed a-pawn. When that job was completed, the king rushed back to aid white's lone king back on the kingside.
• Apparently, N+N+P vs. N+P is a theoretically winning endgame according to Lomonosov Tablebases. My assumption is because white's king is very far away but you may want to train this endgame against a program in order to be confident in a real game.