There is at least one expert on the 2N vs P endgame on this site, MARattigan. And there is one easy advice, the Troitzky line (also see Wikipedia); the rest is complicated. The Troitzky line tells you how far you can permit the pawn to travel before you must block it with a knight and can still win. It's a different place on each file so you must learn and remember the whole line. As MARattigan will tell you there are many exceptions mainly advantageous to the side with the knights. One dutch GM wasted a period of his life trying to improve on the Troitzky line but the tablebases proved that Troitzky was basically right. When a GM stumbles you know it's a difficult ending!
I am pretty sure this is not the first time a topic on two knights versus pawn situation in a chess endgame has been asked, and I have personally seen them in puzzles before.
However, I did not expect myself to actually encounter one such endgame in one of my games. The game ended in a draw (it is my fourth latest draw as of the time of writing). The game is as follows, where I am playing Black in a 2|1 bullet game.
The game is clearly a complete blunderfest as I was living off increment on move 28 onwards with less than 10 seconds to spare.
That said, the position where it ended in a two knights versus pawn situation appeared on move 68 after 68...Nxg6.
When I ran the position on Stockfish, the position was awarded -3.16 (in favour of Black). However, I cross-checked with an endgame tablebase which concluded a draw for the position with best play from both White and Black. And obviously I trust the tablebase more as it is essentially a compilation of all the possible moves arising from such endgames.
Does anyone know how to actually determine from a position whether a two knights versus pawn endgame would result in a win for the side having two knights? And is there some kind of strategy to look out for in both winning positions and drawish positions in such situations?