1...Rxd2 2. Rxd2 Be3+! (the point, as it forces the king to a light square which is important in a couple of moves) 3. Kh1 Bxd2 4. Qxd2 e3 5. Bxc6 Bxc6+ 6. Kg1 exd2 and black is up a piece.
I guess 3. Ke3, f5 (or Bh1 or something), 4. Bb7 are in order to secure a8. I do have to admit I only considered Ke3 and then snapping up the bishop after 3. ... d4+. I didn't realize black should reinforce the bishop or move it to to answer a premature a8=Q which could be answered with d4+ without the king just taking the bishop.
I guess 3. Ke3, f5 (or Bh1 or something), 4. Bb7 are in order to secure a8. I do have to admit I only considered Ke3 and then snapping up the bishop after 3. ... d4+. I didn't realize black should reinforce the bishop or move it to to answer a premature a8=Q which could be answered with d4+ without the king just taking the bishop.
Spoiler:
Nah, Bb7 will always lose to d4+ and Bxb7. I think it has to be 3. Kf4, but the best I can come up with so far is 3.. g5+ 4. Ke5 f6+ 5. Kd4 e5+ 6. Kc5 d4 7. Kb6, with Bb7 incoming. I think we are winning, but it's deep enough for me to start seeing random ghosts and not be able to evaluate properly.. But we HAVE to be winning Q vs 4 pawns, right?!
I'm sure i'm missing something but can't the black King just shuffle between b7 and c7 while the bishop patrols f3 and g2 while also guarding f5 so the knight never has a path to take on g6 or f5 without the bishop capturing back? So the move would be Be4 and I don't see how white makes any progress
1. ... Kc8!!! I think. Amazing. Comps have absolutely no clue what's going on here even though they're quoting depth 30 or something. I keep winning as white against every recommended line.
Was using the ones on http://analysis.cpuchess.com/ and SF/Houdini had no clue even on 45s until way down the lines. Looks like they missed a real draw playing into losing 0.1x evals.
Spoiler:
6. ... c4 in my Nh5 line (plan Be6 and d5) looks like it gets there by a tempo, and if you Nxh7 instead of Kb5, Bc6+ is fine, which was the point of Kb5 to begin with). Hell, even holding that position with the extra pawn is a bitch after 6. Kb5 c4 7. Kb4 Kf7 8. Nxh7 Be6 9. Nf6 d5 10. Kc3 Kf8 11. Kd4 . Not convinced yet. Weird, looking at it again, Houdini has no clue and Stockfish is finding lines fairly fast. Am I at least right for black to move 5?
Was using the ones on http://analysis.cpuchess.com/ and SF/Houdini had no clue even on 45s until way down the lines. Looks like they missed a real draw playing into losing 0.1x evals.
Spoiler:
6. ... c4 in my Nh5 line (plan Be6 and d5) looks like it gets there by a tempo, and if you Nxh7 instead of Kb5, Bc6+ is fine, which was the point of Kb5 to begin with). Hell, even holding that position with the extra pawn is a bitch after 6. Kb5 c4 7. Kb4 Kf7 8. Nxh7 Be6 9. Nf6 d5 10. Kc3 Kf8 11. Kd4 . Not convinced yet. Weird, looking at it again, Houdini has no clue and Stockfish is finding lines fairly fast. Am I at least right for black to move 5?
Spoiler:
I was treating the critical endgame (N vs B with f and g pawns each) as drawn. It looks tricky but Black draws if he keeps the king out of d6 (computer backs me up here). So instead of 11...c4, 11...Kg7 12. Kxc5 Kf8 and 13...Ke7. The knight can't win a pawn on its own, the White king can't enter, and with his king this far west, Black also holds pretty much every pawn ending.
Two hints:
1: there is a more thematic refutation to 1...Be4
2: Your line would definitely be winning for White if Black didn't have the move you mention,
Spoiler:
5...Ke6!
If you understand that, that's pretty much key to solving the rest of the position
Lame, I kept winning all of those positions (including K on e7), but the evals quit too soon so it kept walking back into the corner. Can win the g-pawn by force with Ne5 ideas and get something like Kf6 Ng6 f4 g5 vs Ke4 f5 Bb3 but there's no win from there. Guess Be4 Kb5 and then there's no silliness after Kc7 Nf1 with the same ideas.
That's pretty. Since 1.Kd7?? Rf4 is a draw due to, e.g. 2.c8=Q Rd4-c4xc8 1. Kd5! is the way, and White can walk down the rook to prevent the skewer, e.g. 1...Rf5+ 2.Kd4 Rf4+ 3.Kd3 Rf3+ 4.Kc2 Rf2+ and now back 5.Kb3 Rf3+ 6.Kb4 Rf4+ etc. After Kb7 the king is clear of annoying checks.