two knights vs rook

You should've taken the rook with the fork instead of the bishop. I like my chances with single knight vs single bishop much better.

Knight c6 is the best, because it limits rook activity on backrank and seventh rank, so covering e7 and d8 squares and protecting the a7 pawn, this will make it hard for to attack your pawns; Your other knight can come to e7, they protect and coordinated, this will allow you to advance the king in the center. It be impossible to lose this endgame but loose knights can be a target.

You lose because you fail to protect essential squares, one them being a e5, e7 d8 squares, knights need the aid of the king and so you can protect and centralize your king, so setup will be knight onn c6 and e7 and king going to f6. You lost your d5 pawn in two moves thse squares pointed out weren't cover by you knights. Endgame you must make a plan and where you going to place your pieces. This one I immediate realize you must limit the rook scope. 

In my opinion, you promoted a pawn because you had done the good work of getting up 2 pawns not because you had knights, so he or she didn't have time to win all three by the time you promoted.

Generally depending on the position the game is a draw,however as you mentioned their were still pawns on the boards which leads me to think that it was about equal.Although i do not know the position which you are referring to i think if you are able to eliminate all the pawns and have a rook vs 2 knights it is a draw.

Never mind,white is completely winning in the game shown above,2 knights vs a rook plus 2 extra pawns equals an easy victory.

Check out this #chess game: ChandraUmeshRaavi vs Alpha139 - https://www.chess.com/daily/game/243647564 I recently played such an end game. I was somehow able to fork and take the rook.

It was no accident IMO; White's weak kingside pawns (doubled and isolated) are such a liability that the knights have an easy time here.

Fascinating

Don't know much about this endgame but my understanding is that generally 2 knights have a slight advantage, but should be drawn if both sides are active.

RPvNNP is peculiar in that the rook could be given up for the pawn, or less commonly, the knight, if the side w/ 2N has a rook pawn they may be able to draw (with the fortresses with KNPvK with rook pawn). Knowing when to sac rook for pawn in an otherwise lost position requires a bit of 2NvP endgame knowledge, but the chances are virtually nonexistent and I wouldn't even recommend preparing for this. If the side with the rook is winning and they sac rook for pawn to promote their pawn... well then we have a Qv2N ending which is even rarer.

On the other side, giving up a knight for a pawn almost always ends up in a drawn position, and as there are 2 knights this shouldn't be too hard usually.

I guess the only practical winning chances come when either side has an extra pawn, but even then, in RPvNNPP if you can trade pawns and sac rook for last pawn you got the draw. Maybe there could be a knight for pawn sac, and then a RvNPP ending might have good winning chances for the knight. Or even a rook for knight sac if the last remaining pawn is a rook pawn and a fortress can be maIde.

In RPPvNNP you could either sac both knights for both pawns (if the resulting RvP position is drawn of course), trade pawns and sac knight for pawn for a hopefully easy RvN draw, or just... i don't even know...

All this is just irrelevant I think at this point

Don't resign when you could hold the draw, just like you don't want to agree a draw when you have all the winning chances, like in this position:

30. (black) Rd7 was better. This way, you keep a rook on the board and can control the vertical files better. White has no more dangerous counterplay here. If this move was played, engine eval = -3.31. In almost all cases (even if the rook seems to be active), two pieces (even two knights) are a lot better than a rook.

After 23. (white) Rae1, black is much better despite the material equality and the seemingly active rooks (according to Stockfish). The knights are blocking the rooks activity on the open file. (Even in the early endgame, two pieces usually dominate over a rook and pawn. In one game, I had couple endgames in which I had a rook and two bishops vs the opponent's two rooks and two pawns; the engine evaluated my position as slightly better despite being down 1 point in material. This is mainly because I have an extra piece and my pieces were coordinated.)

In the endgame a lone rook often outweighs two minor pieces. They lack a coordinator. So adding a rook to both sides (or a Queen) which acts as coordinator the two minor pieces will prevail. For the minor pieces: keep heavy pieces, for the rook party exchange!

Classic example, the rook is quite powerful:

https://www.chessgames.com/perl/chessgame?gid=1552975