ECM 1381

[jdart]

This position is in the "ecmgcp" test suite, which is a subset of ECM (Encyclopedia of Chess Middlegames). It looks like there are multiple high-scoring moves, and the target move (Qb2) is not actually scoring highest (analysis by Stockfish 17, 30 cores on a Ryzen Threadripper, 200 seconds/move):

2b5/2qrrpk1/5Rp1/2p4p/1pB1PR1P/1P1P2P1/5Q1K/8 w - - bm Qb2; id "ECM.1381";

depth 36 seldepth 75 multipv 1 score PovScore(Cp(+655), WHITE) nodes 3619302600 nps 18096422 hashfull 529 tbhits 0 time 200.001 pv Kg2 Kh7 Bxf7 Qe5 Bxg6+ Kg7 R6f5 Qd4 Rg5 Qxf2+ Kxf2 Rd8 Bf7+ Kh6 Rxh5+ Kg7 Rg5+ Kh7 Ke3 Ba6 Bd5 Bc8 Rh5+ Kg7 Bc4 Be6 Bxe6 Rxe6 Rg4+ Kf7 Rxc5 Rdd6 Rf5+ Ke7 Rg7+ Ke8 Rg8+ Ke7 h5 Rb6 Rfg5 Ra6 R5g7+ Kd6 Kd4 Ra1 Rg6 Ke7 Rxe6+ Kxe6 g4 Rb1
depth 35 seldepth 72 multipv 2 score PovScore(Cp(+534), WHITE) nodes 3619302600 nps 18096422 hashfull 529 tbhits 0 time 200.001 pv Qb2 Rd4 Rxf7+ Rxf7 Rxf7+ Qxf7 Bxf7 Kxf7 Qc2 Be6 Kg2 Rd8 Kf2 Rc8 Qa2 Rc7 Qa5 Rb7 Qxc5 Bxb3 Qc6 Re7 Qb5 Be6 Qxb4 g5 Ke3 Bh3 Qd6 gxh4 gxh4 Re6 Qc7+ Kf8
depth 35 seldepth 78 multipv 3 score PovScore(Cp(+500), WHITE) nodes 3619302600 nps 18096422 hashfull 529 tbhits 0 time 200.001 pv Qf1 Qe5 Rxf7+ Kh6 Rf8 Rc7 R4f6 Kh7 Qf2 Bg4 Bf7 Red7 Bxg6+ Kg7 Be8 Rd6 R6f7+ Rxf7 Rxf7+ Kh8 Bb5 Qd4 Bc4 Qxf2+ Rxf2 Rd8 Rf7 Be2 Rc7 Bxd3 Bxd3 Rxd3 Rxc5 Rxb3 Rxh5+ Kg7 Rb5 Rd3

[peter]

After some Forward- Backward with 30 threads of 16x3.5GHz CPU:

[d]2b5/2qrrpk1/5Rp1/2p4p/1pB1PR1P/1P1P2P1/5Q1K/8 w - - 0 1

Analysis by SlowChess Blitz 2.9 avx2:

1.Kg2 Kf8 2.g4 hxg4 3.Txg6 Td4 4.Lxf7 Txf7 5.Txf7+ Dxf7 6.Tf6 Dxf6 7.Dxf6+ Ke8 8.De5+ Kd8 9.Dg5+ Kd7 10.Dxc5 Td6 11.Dxb4 Kc7 12.Dc3+ Tc6 13.Dg7+ Ld7 14.h5 Tc2+ 15.Kg3 Tc1 16.h6 Tg1+ 17.Kh2 Tf1 18.h7 Tf3 19.Kg2 Th3 20.h8D Txh8 21.Dxh8 Kc6 22.Kg3 Kc5 23.Dc3+ Kb6 24.Kf4 g3 25.Kxg3 Lc6 26.Kf4 Kc7 27.Dd4 Lb7 28.Ke5 La8 29.Dd6+ Kc8 30.Df8+ Kb7 31.Dc5 Kb8 32.Db6+ Lb7 33.Kd6 Ka8 34.Kc7 Lc8 35.Db8#
Tiefe: 36/81 00:01:25 2078MN, tb=13445486
+- (#35)

And after changing to MultiPV= with full hash:

1. +- (#35): 1.Kg2 Kf8 2.g4 hxg4 3.Txg6 Td4 4.Lxf7 Txf7 5.Txf7+ Dxf7 6.Tf6 Dxf6 7.Dxf6+ Ke8 8.De5+ Kd8 9.Dg5+ Kd7 10.Dxc5 Td6 11.Dxb4 Kc6 12.Dc3+ Kb7 13.h5 Td7 14.h6 Th7 15.De5 Kc6 16.Df6+ Kc7 17.Dg6 Td7 18.Dxg4 Txd3 19.Dg7+ Td7 20.Dxd7+ Lxd7 21.h7 Lc6 22.h8D Lxe4+ 23.Kg3 Ld5 24.Kf4 Kc6 25.Dc3+ Kd6 26.Dd4 Kc6 27.Ke5 Lh1 28.Dc3+ Kd7 29.Dh3+ Kc6 30.Dxh1+ Kc7 31.Da8 Kb6 32.Kd6 Kb5 33.Db8+ Ka5 34.Kc7 Ka6 35.Db6#
2. +- (9.43): 1.Db2 De5 2.Txf7+ Kh6 3.Dxe5 Txe5 4.Tf8 Lb7 5.T4f6 Tee7 6.Tb6 Tc7 7.Th8+ Th7 8.Txh7+ Txh7 9.Kg2 Lc8 10.Kf3 Ta7 11.Kf4 Ta1 12.Ke5 Tg1 13.Tb8 Lg4 14.Kd6 Txg3 15.e5 Kg7 16.e6 Lxe6 17.Lxe6 Txd3+ 18.Kxc5 Kf6 19.Ld5 g5 20.Lc4 Tg3 21.hxg5+ Txg5+ 22.Kxb4 Ke7 23.Tb5 Txb5+ 24.Kxb5 Kd6 25.Kb6 h4

Of course at least this second one move would need extra Forward- Backward to see, if DTM maybe would be about as long, anyhow I'd say this position is more a matter of search but a one of eval and probably having not very much longer distances to mate with other but these ones first moves, it won't be a position of clear unique single best move in sense of an engine test position, as long as exact DTM of the candidate moves aren't known. Don't have the hardware- time and interactive time at the moment to find out for sure, regards

[Eelco de Groot]

Just a little independent analysis;

Rebel (EAS of course) would deviate (a little) (in the Qb2 line) with 6.Rg8 (as does Sun SE). That is if you give the position from Slowchess after the moves 1 to 5, ( 1.Db2 De5 2.Txf7+ Kh6 3.Dxe5 Txe5 4.Tf8 Lb7 5.T4f6 Tee7 ) but both more or less agree with those moves, I think.

First analysis was

[d]5R2/1b1rr3/5Rpk/2p4p/1pB1P2P/1P1P2P1/7K/8 w - -

Engine: Rebel EAS 3.0 (1024 MB)
gemaakt door Ed Schroder & Chris Whittington

41/82 55:27 +8.83 6.Tg8 Tg7 7.Th8+ Th7 8.Tb8 Tc7 9.Kg2 Tc8
10.Txc8 Lxc8 11.Kf3 Tc7 12.Kf4 Lh3
13.Ke5 Kg7 14.Tb6 g5 15.Kd6 Tc8
16.e5 Tf8 17.Kxc5 Tf3 18.Tb7+ Kh6
19.hxg5+ (2.177.216.571) 654

Restart of Rebel by inserting the line in the pgn (bug in Shredder GUI) now follows, for instance, 6.Rg8 Rg7 7.Rb8 Rc7 8.Kg2 Bc8 9.Kf3 Rge7 10.Kf4 Re8 11.Rbb6 Rg7 12.RRbc6 Bg4 13.Rxc5. Going to this position


[d]4r3/6r1/5Rpk/2R4p/1pB1PKbP/1P1P2P1/8/8 b - -

Engine: Rebel EAS 3.0 (1024 MB)
gemaakt door Ed Schroder & Chris Whittington (C) 2026


36/71 67:26 +10.67 13...Td8 14.Tcc6 Ld1 15.Tb6 Lc2
16.Ke3 Ta8 17.Tf2 Ld1 18.Txb4 Lg4
19.Ta4 Txa4 20.bxa4 Ta7 21.Ta2 Ta5
22.Lb5 g5 23.d4 Kg7 24.d5 Kf6 25.Kd4 Ld1
26.Ta1 gxh4 (2.583.760.369) 638

It now depends a little if you declare this (after 13.Rxc5) a won position, or want to be demonstrated how many moves to Mate. Back in the day, (of ECM 1381) even with the strongest hardware, IMO, it was nearly impossible to find that Mate in 35 (of the other line) I think. Simply beyond capabilities then. At the same time I think many players would fold after 13.Rxc5 being two connected passers behind, with Black. Unless a lot of money or titles are at stake. Like in the Candidates tournament right now. Then you'd play on with Black. I think. (well maybe not even then...) So solution is okay, given the times, I'd say.

[Ajedrecista]

Hello:

A blast from the past: searching the FEN retrieved a result from the old CCC archive (hoping that its unaivalability is temporary) that was luckily archived into the WayBack Machine:

Re: A proposed WAC replacement for testing

I ran for two minutes per position in K=2 mode, and here is what I came up with.

It's hard to decide exactly what constitutes a cook. In purest sense, a cook should be a mate that is no slower than the key allows.

I think that positions with approximately the same large score could be confusing, and I found a lot of those.

I would throw these out, to be safe.

Your mileage may vary.

[...]

Id: ECM.1381
Fen: 2b5/2qrrpk1/5Rp1/2p4p/1pB1PR1P/1P1P2P1/5Q1K/8 w - - 0 1
The key is Qb2. I have Kg2 as over +2 also, with a higher score at the end.

[...]

Red letters added by me. The post is from September of 2001, almost 25 years ago!

Regards from Spain.

Ajedrecista.

[Jouni]

I think up to 30 % of pre-computer puzzles are busted! And also 5-10% of puzzles before strong computer programs. One famous example from BT2450 suite:

[d]3r2k1/p2r2p1/1p1B2Pp/4PQ1P/2b1p3/P3P3/7K/8 w - - 0 1
Solution 1.e6. Really? 1. Bb4 is much better move.