11

In this position, Stockfish gives black (me) -8.1 and recommends BxN.

8/5p2/4pkpp/p1n5/Pp6/1Bb1PP2/2PNK2P/8 b - - 7 28

Instead I played Kf5 and the evaluation plummets to -2.1.

My understanding was that generally bishops are preferred over knights in endgames, so I'm trying to understand what is different about this position. Looking at Stockfish's line, it looks like somehow this is useful for dragging the king away from protecting the pawn on H2, which Black's king then goes after.

Is there a useful principle to learn from this?

Improve this question
1
  • When analyzing with an engine, don't just stick to the next move. In the alternative Bxd2, the engine wants to trade every piece on the board to get a winning pawn endgame.
    – David
    Jun 28 at 15:29

3 Answers 3

Reset to default
12

Easy. Exchange BOTH pairs of light officers (BxN, then NxB) and you have an almost trivial won pure king/pawn endgame. General Principle: If you are material up in the endgame, exchange officers. If down, pawns. (As usual, exceptions abound.)

Improve this answer
12

After Kf5 white replies with Nc4 and the only way to avoid losing the a pawn is to play the knight back to the poor square b7 and it is very difficult to see how black can make progress. If the knights are ever exchanged then the position becomes a trivially drawn opposite coloured bishop endgame.

[fen "8/5p2/4pkpp/p1n5/Pp6/1Bb1PP2/2PNK2P/8 b - - 0 1"]

1...Kf5 2. Nc4 Nb7

On the other hand, if black exchanges off the bishop for the knight by Bxd2 then black is left with a good knight versus a bad bishop with a potential distant passed pawn on the king side. (Note that if the white bishop moves then that immediately loses the a pawn.) Black will play moves like g5, Ke5, f5, g4.

Endgames are all about calculation and so black should also calculate carefully what happens if the other minor piece is also exchanged by Nxb3. This opens up the route for white to play a sequence of moves like Kd3-c4-b5, Kxa5 and queen the a pawn. Can white win via this route? The answer is not quite. With correct play black queens the h pawn one move before white can queen the a pawn and as they are on the same diagonal black's queen covers a8.

[fen "8/5p2/4pkpp/p1n5/Pp6/1Bb1PP2/2PNK2P/8 b - - 0 1"]

1...Bxd2 2. Kxd2 Nxb3 3. cxb3 Ke5 4. Kd3 f5 5. Kc4 g5 6. Kb5 g4 7. fxg4 fxg4 8. Kxa5 h5 9. Kxb4 h4 10. a5 g3 11. hxg3 h3! (11...hxg3? {only draws}) 12. a6 h2 13. a7 h1=Q

Is there a useful principle to learn from this?

Calculation is key in endgames. If you watch the top players they don't rush in and play the obvious move when they think they have a winning endgame. They take the time to calculate the position through to the win and only then play the move.

Improve this answer
3
  • 4...Kd5 seems safer than going for the pawn race.
    – eyeballfrog
    Jun 23 at 14:25
  • Instead of 7 fxg4 white could play f4+ and very likely getting the queen first then getting to more complicated endgame (black king would probably go Kf6, otherwise (Ke4) white will get queen with check.
    – Adam Tucholski
    Jun 25 at 9:17
  • 2
    @AdamTucholski No. After 7. f4+ Kd5 8. Kxa5 Kc5 9. Ka6 Kc6 the white king is never going to succeed in getting out of the way of the a pawn and black is going to queen and win.
    – Brian Towers ♦
    Jun 25 at 14:55
2

If the engine goes -8.1, it found a clear win, so subtle strategic considerations are not relevant anymore for the engine. But for the human, they may be.

The pawn endgame is less clear, far from trivial. Pawn endgames can be extremely complex and I would not transition unless I am 100% sure that it is winning. Transitioning here is an unnecessary gamble. In this position I believe you maintain some serious advantages if you go Bxd2 and then keep your knight on c5.

The knight on c5 is cutting the king off, is preventing c3 or c4, and is threatening to capture on a4 anytime the bishop moves. With that in mind, black can increase the pressure by going for the h-pawn with his king. If white choses to guard h2, then he ends up in a totally passive position where he cant move his king or bishop, and will soon run out of moves. So basically, white will have to let go of another pawn, sooner rather than later.

If there is one lesson to be learned from this position i believe its this:

Don't just follow strategic dogmas, but examine if they really apply.

Improve this answer
6
  • 1
    Why do you say the pawn ending is not a simple win? It sure looks like so to me, since Black is up a pawn, has a pawn majority (no doubled pawns, and the queenside pawns are also far-advanced so White has to always be mindful of a ...Kxb3 maneuver.
    – Allure
    Jun 23 at 0:58
  • Its objectively won, but I am seeing this from a practical perspective. Simplification doesn't necessarily mean its simpler to play. Pawn endgames can be notoriously tricky, one miscalculation and the win is out of the window. I am not saying they should be avoided, but given that there is a choice, go with the simplest and clearest winning plan. For me, in this position, its keeping the knight on the board.
    – user35042
    Jun 23 at 11:27
  • I had no trouble beating Stockfish in the pure pawn endgame. It seems really simple. 1...Bxd2 2. Kxd2 Nxb3+ 3. cxb3 Ke5 4. Kd3 Kd5. Black has the opposition, a pawn majority on the kingside, and White cannot allow ...Kxb3. But White has to move the king, or Black's kingside majority pushes through. It doesn't look hard to play at all.
    – Allure
    Jun 23 at 11:57
  • Sorry but the topic of discussion is not whether the pawn endgame is easy to play, or whether you have beaten Stockfish, but how the position is evaluated strategically and what lessons could be drawn from it. I explained how I would think about it and why I personally would avoid the pawn endgame. Not sure what your point is.
    – user35042
    Jun 24 at 12:36
  • 1
    The point is you have written "The pawn endgame is less clear, far from trivial". Since that is in your answer, the question of whether the pawn endgame really is far from trivial is relevant.
    – Allure
    Jun 24 at 13:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.