[nalihbyw] Bongcloud endgame

the Bongcloud is a chess opening popularized by GM Nakamura.

1.e4 e5

standard king's pawn opening, though similar ideas are possible after 1.d4 d5.

2.Ke2

this is the Bongcloud Attack.  traditional untraditional openings for white focus on the first move, e.g., 1.h4 .  however, because of first-move advantage, white cannot do too much more than merely equalize on the first move, which is not that interesting.  the genius of the Bongcloud is in its deferring of inhalation until the second move, when this far more interesting king move is possible.

2...Ke7

i first heard of this defense from GM Carlsen.  he proposed the continuation 3.Ke1 Ke8 which leads to crazy complications ("no-castling chess"), a variation also recently investigated in depth by GM Kramnik and the AlphaZero team.  (alternatively, 3.Qe1 Qe8 4.Kd1 Kd8 is an interesting try, inspired by Carlsen's Norwegian variations of the Bongcloud which involve swapping king and queen through the queen making a circuit across the 4th rank.)

3.Kd3

white commits to the Bongcloud Attack actually being an attack.  white might have plans of Kc4, Kd5, Kxe5 picking off black's valuable center pawn and eyeing the soon-to-be unguarded f7 pawn, a classic weakness in black's camp.

the text move is preferable over 3.Kf3 because controlling the center is fundamentally important.  also, many checkmates involve trapping the king against the edge of the board, so we keep the king as far as possible from those danger zones.

3...Kd6

black defends thematically.  f7 is left unguarded.

4. Kc4

with the kings established as attacking pieces, we are now officially in the endgame.

4...Kc6!

one of the most important principles of the endgame is opposition.  black breaks symmetry (albeit not by choice, illegal is 4...Kc5), seemingly playing a somewhat passive move to counter white's aggression.  however, black gains opposition.  now, white's king cannot easily invade black's side of the board: any sideways move by white, trying to outflank, can be countered with a corresponding sideways move by black.

Stockfish evaluates the position as even (depth 62).  however, computers are not trained on positions like this, so who knows if the evaluation can be trusted.

Stockfish ee3f7b6b6e1a1051b32cedb38ac89b3458ff4ab2
nn-62ef826d1a6d.nnue
make clean profile-build ARCH=x86-64-avx512
gcc 10.2.0
Threads 1 (default)
Use NNUE true
Hash 20000 (20GB)
position startpos moves e2e4 e7e5 e1e2 e8e7 e2d3 e7d6 d3c4 d6c6

info depth 62 seldepth 76 multipv 1 score cp 0 nodes 516497810720 nps 1779912 hashfull 1000 tbhits 0 time 290181556 pv c4b3 b7b6 f2f4 e5f4 d2d4 c6b7 c1f4 f7f5 e4f5 d7d5 b1c3 c8f5 a2a3 b8c6 f1b5 g8f6 g1f3 f8d6 b5c6 b7c6 f4d6 d8d6 f3e5 c6b7 h1f1 f5g6 b3a2 a8e8 d1d2 f6e4 c3e4 d5e4 d2e3 c7c5 e5g6 c5d4 f1f7 b7a8 e3f4 d6d5 b2b3 d5c6 f7c7 c6g6 a1f1 g6e6 f4f7 e6f7 f1f7 e4e3 c7a7 a8b8 f7b7 b8c8 b7c7 c8b8 a7b7 b8a8 a2b1 e3e2 b7a7 a8b8 c7b7 b8c8 b1a2 e2e1q b7c7 c8b8 a7b7 b8a8 b7a7