Rook vs two knights endgame analysis help

Sort:
Misteribel

Against a bot I ended up with the following position, which I thought was a draw. I won, but that was just because the training bot was making mistakes (it was an 1100 bot, so I could win).

In this position I was at a loss how to force a win. Apparently the next two moves by white are critical. I didn’t find them. Does anybody have tips for finding a way out against rook with two minor pieces? I originally went for b4, which is apparently a draw. Is there a tactic to apply here or is it a matter of experience? Is the winning move obvious to the more experienced eye and if so, why? (I only found the proper tactic using SmallFish as the analysis of chess.com seemed weird, it kept losing).

reasonably_happy

b4 is a blunder, because h6 and your knight is trapped. Even if it's a draw after this, it looks extremely tricky to hold and black is one playing for the win. The only sensible first move to me is Kd3, you are now safe from all black's threats and in general obviously better. Then you start pushing your passer, he's bound to break eventually. Just look out for him sacking the rook for your last pawn, because you'll end up with KNN vs K which is a draw.

Misteribel

@reasonably_happy, thanks for your response! I see now why Kd3 is the only correct move. Meanwhile I played against max computer to try if I could get out of here. After some tries I managed to end up just like you predicted, Kd3, Rb7; Ne4, Rxb2; Nf6+, Kc5; Nd4, Kb4; Nxe6, Rb3+; Kd4, Rxh3; which gives:

 

According to tablebases, this is winning and white can keep his pawn. But then I blundered again and ended up with NN against king. This requires 112 plies to win, which is a few moves too many! Since black has a pawn, I tried it anyway (it’s on the Trozky line).

 

Another couple of tries later it turned out that the max computer strength on chess.com apparently doesn’t use tablebases. With some outside help I managed to get to this, which is trivially winning (ok, not that trivial, but within 20 moves and another practice round for me).

 


This was fun and a bit weird. I guess better moves can be made by either party, a bit surprised at the logic of Stockfish here. When playing that last position, you’d imagine Stockfish, at highest settings, would make the right moves, trying to draw, but instead it was over in far fewer than 20.