Vinvin wrote: ↑Mon Apr 27, 2020 2:09 am
Stockfish (and some derivatives) still have troubles with the mainline here :
Initial position : r3kb1r/1p4pp/2q1pp2/1p2Q3/8/2P5/PP2B1PP/R4R1K w kq - 0 25
Mainline : 1. Rxf6 gxf6 2. Qxf6 Rg8 3. Bh5+ Rg6 4. Bxg6+ hxg6 5. Rf1 Kd7 6. Qf7+ Kd6 7. Rd1+ Ke5 8. h3 Rc8 9. Re1+ Kd5 10. b4 b6 11. c4+ bxc4 12. Rd1+ Ke5 13. Qxg6 Qe4 14. Qg5+ Qf5 ...
1) find 8.h3! with a big eval (possible transposition : 8. Re1+ Kd5 9. Rd1+ Ke5 10. h3)
2) find 10.b4! with a big eval
3) find 11.c4+! with a big eval (possible transposition : 11.Rd1+ Ke5 12.Re1+ Kd5 13.c4+)
SF seems to be better here then all the derivates I tried so far.
2r2b2/1p3Q2/2q1p1p1/1p1k4/8/2P4P/PP4P1/4R2K w - - 0 1
Analysis by Brainfish 200420 64 POPCNT:
Code: Select all
34.Rd1+ Ke5 35.Re1+ Kd5 36.b4 g5 37.a4 bxa4 38.b5 Qe8 39.Qxb7+ Kc5 40.c4 a3 41.Qa7+ Kxc4 42.b6 Bc5 43.Qa6+ Kd5 44.b7 Rb8 45.Rd1+ Ke5 46.Qd3 Kf6 47.Qc3+ Kg6 48.Qxc5 Rxb7 49.Qxa3 Qb5 50.Re1 Qf5 51.Qa2 Rb6 52.Qa1 Qf6 53.Qa7 e5 54.Qd7 Kh6 55.Qd1 Kg7 56.Qc2 Rb7 57.Qc3 Rf7 58.Qxe5 Qxe5 59.Rxe5 Kf6 60.Re2 Rc7
= (0.15) Depth: 39/64 00:00:45 1188MN
34.Rd1+
= (0.25 ++) Depth: 40/51 00:00:47 1236MN
34.Rd1+
+/= (0.35 ++) Depth: 40/51 00:00:48 1241MN
34.Rd1+
+/= (0.59 ++) Depth: 40/56 00:00:52 1346MN
34.Rd1+
+/- (0.79 ++) Depth: 40/56 00:00:52 1348MN
34.Rd1+
+/- (1.06 ++) Depth: 40/56 00:00:52 1352MN
34.Rd1+
+/- (1.42 ++) Depth: 40/76 00:00:54 1396MN
34.b4 b6 35.c4+ bxc4 36.Rd1+ Ke5 37.Qxg6 Qe4 38.Qg5+ Qf5 39.Qh4 Qf4 40.Re1+ Kf5 41.Qh7+ Kg5 42.Qg8+ Kh5 43.Qxe6 Bd6 44.Qd5+ Kh6 45.Re6+ Kg7 46.Rxd6 Qc1+ 47.Kh2 Qf4+ 48.g3 Qf2+ 49.Qg2 Qxg2+ 50.Kxg2 c3 51.Rd1 Kf6 52.Kf3 Ke5 53.Ke2 Ke4 54.a4 c2 55.Rc1 Rc3 56.g4 Rxh3 57.Rxc2 Kd5 58.Rd2+ Ke5 59.Rd8 Rb3 60.Re8+ Kd5 61.g5 Rxb4 62.Kf3 Rxa4 63.g6 Ra7 64.Rb8 Kc5 65.Ke4 b5 66.Kf5 b4 67.Kf6
+/- (1.45) Depth: 40/79 00:01:02 1569MN
...
34.b4 b6 35.c4+ bxc4 36.Rd1+ Ke5 37.Qxg6 Qe4 38.Qg5+ Qf5 39.Qh4 Qf4 40.Re1+ Kf5 41.Qh7+ Kg5 42.Qg8+ Kh5 43.Qxe6 Bd6 44.Qd5+ Kh6 45.Re6+ Kg7 46.Rxd6 Qc1+ 47.Kh2 Qf4+ 48.g3 Qf2+ 49.Qg2 Qxg2+ 50.Kxg2 c3 51.Rd1 Kf6 52.Kf3 Ke5 53.Ke2 Ke4 54.Rf1 Ra8 55.Rf4+ Kd5 56.Kd3 Rxa2 57.Rf5+ Kc6 58.g4 Rh2 59.Kxc3 Rxh3+ 60.Kc4 Rh6 61.Kd4 Rh1 62.b5+ Kd6 63.Rf6+ Kc7 64.g5 Rb1 65.g6
+- (1.96) Depth: 41/73 00:01:12 1826MN
...
34.Rd1+ Ke5 35.Re1+ Kd5 36.b4 b6 37.c4+ bxc4 38.Rd1+ Ke5 39.Qxg6 Qe4 40.Qg5+ Qf5 41.Qh4 Qf4 42.Re1+ Kf5 43.Qh7+ Kg5 44.Qg8+ Kh5 45.Qxe6 Bd6 46.Qd5+ Kh6 47.Re6+ Kg7 48.Rxd6 Qf1+ 49.Kh2 Qf4+ 50.g3 Qf2+ 51.Qg2 Qxg2+ 52.Kxg2 c3 53.Rd1 Kf6 54.Kf3 Ke5 55.Ke2 Ke4 56.Rf1 Ra8 57.Rf4+ Kd5 58.Kd3 Rxa2 59.Rf5+ Ke6 60.Rg5 Rh2 61.Kxc3 Rxh3 62.Kc4 Kf6 63.Rg8 Kf7 64.Rc8 Rxg3 65.Kb5 Rg4 66.Rc4 Rg6 67.Rc6 Rg4 68.Rxb6 Ke7 69.Rc6 Kd7 70.Rc1 Kd6 71.Ka5 Rg3 72.b5 Rb3 73.Ka6
+- (2.46) Depth: 42/90 00:01:31 2279MN
...
34.b4 b6 35.c4+ bxc4 36.Rd1+ Ke5 37.Qxg6 Qe4 38.Qg5+ Qf5 39.Qh4 Qf4 40.Re1+ Kf5 41.Qh7+ Kg5 42.Qg8+ Kh5 43.Qxe6 Bd6 44.Qd5+ Kh6 45.Re6+ Kg7 46.Rxd6 Qc1+ 47.Kh2 Qf4+ 48.g3 Qf2+ 49.Qg2 Qxg2+ 50.Kxg2 c3 51.Rd1 c2 52.Rc1 Kf6 53.Kf3 Ke5 54.Ke2 Rc3 55.g4 Kd5 56.Kd2 Rxh3 57.Rg1 Ra3 58.g5 c1R 59.Kxc1 Rxa2 60.Rg4 Ra7 61.g6 Rg7 62.Kd2 Ke5 63.Kd3 b5 64.Rg1 Rd7+ 65.Ke3 Rg7 66.Rg5+ Kf6 67.Rxb5 Kxg6 68.Kd4 Rc7 69.Kd5 Kf6 70.Rb6+ Ke7 71.Rc6 Rd7+ 72.Kc5
+- (2.80) Depth: 44/100 00:02:16 3400MN
At three trials time to solution (+eval) SF was < or at 1 minute at all of them.
As John showed above, Rd1 with b4 in ouput- line two moves later, simply repeating position to be solved once (or even two times), has to be counted as solved too, if output and eval is right.
Position just before 35.(11.) c4 is of course to be solved for one move more easily, yet the derivates still take more time than SF dev. takes for b4 at my trials.
LC0 needs about a minute for c4, broke off after about 5 miuntes not finding b4 at 2080ti.
Peter.