18...Ne3, by forcing an exchange of minor pieces, would have left Black with a strong bishop for knight on a fairly open board.

Why do you think that this check is dangerous? You can always play Kg1 and all is well. Not that 21.g3 is necessarily bad, but it is not the only good option here.

The check wasn't really dangerous. But I suspect bleddy saw the potential for 21. ... Qh4+ 22. g3 Qxh2+. But instead of 22. g3, white could have always played Kg1...and everything is safe.

OK, but I rather like 22.Nc3, with a view to forcing off the blocking Black knight.

This looks OK IMHO, no need to force things at this point.

Not only is this OK, Ne4 was actually best and either wins the c6 pawn or gains substantial position. White should have been able to force a win at this point under best play. **** On the contrary, Trading the knights is the bad idea. There isn't much to gain from that at all.

Exchanging knights would have given White a useful Q-side pawn majority, but Black's central majority would go far to counterbalance it. I agree with the other commentators: the move played wins material and offers thereby much better and more varied winning chances.

after 25.Nxd5, doesn't white gain a pawn with 26.Rc6 ?

This is more than just palpable. This is game winning position. You just didn't know how to put the final nail into the coffin.

You took the wrong pawn. The c pawn was much more important. Also a white rook on c6 adds significant pressure upon Black's position. It is a house of cards ready to fall. Invasion was the answer.

Yet the move played is very plausible. The c-pawn sits isolated on a file open to the c1-rook, and is attacked twice with no defender. Surely it must soon fall? I know I would look first at the a6-capture. But I would probably have chosen the alternate one as lighttotheright indicates. By taking the c-pawn at once, White would have kept the knight out of b4, leaving the enemy with fewer opportunities. When assessing choices such as the one White faced here, one has also to assess the opponent's chances of developing counterplay.

Even with the wrong pawn taken and this black knight move, white should still be able to force a win...it just won't be quite as easy to force.