Looking for a mechanical solution to 11 Epd problems

[Dann Corbit]

These are from Vincent Lejeune's recent set of 115 chess problems. I have mechanical (computer generated) solutions to all but 11 of them. Some are incredibly difficult.

For instance, with this problem:
http://www.yacpdb.org/?id=340530
I tried creation of all the positions to the solution and then analyzing them backwards to feed the hash table. Eventually, the solution gets so far away that the engines get lost anyway.

If anyone has mechanical solutions to any of these problems, I would be very glad to receive them.

8/4ppkp/3p2p1/2pP1PP1/2P1P1P1/p7/K3B3/2b5 b - - acd 60; acs 1215; bm Bxg5; c0 "Bxg5 definitely wins. Does f6 win faster?"; c1 "diff=08"; c3 "f6"; c4 "(Dvor) 163"; ce 549; id "Dvoretsky.126"; pm Bxg5; pv Bxg5 fxg6 fxg6 Kxa3 Kf6 Kb2 Ke5 Kc2 Kxe4 Kd1 Bh4 Kd2 Kd4 Bf1 Bg3 Be2 Bf2 Bf1 Bh4 Be2 Bg3 Kc2 Ke3 Bd1 Be5 Kb3 Kd3 Bf3 Bf6 Bd1 Kd2 Bf3 Bd4 Ka3 Ke3 Bd1 Kf4 Be2 h5 gxh5 g5 Bf1 Kf3 h6 g4 h7 g3 Kb3 g2 Bxg2+ Kxg2 Kc2 Bf6 Kd3 Kf3 Kd2 Ke4 Kd1 Kd3 Ke1 Kxc4 Ke2 Kxd5;
8/1p2b3/6k1/5p2/p2NbP1p/P1B1P1p1/1P5P/6K1 w - - acd 65; acs 2720; bm Be1; c0 "Event:Smirin vs Computers; Site:KC-CB INT; Date:2002.04.24; Round: 6; Smirin, Ilia vs Hiarcs 8"; c1 "diff=10"; c3 "h3"; ce -26; pm Be1; pv Be1 gxh2+ Kxh2 Kh5 Kh3 Bd3 Nf3 Bf1+ Kh2 Bc5 Bd2 Bb6 Ne5 Bb5 Kg2 Bc5 Kf3 Be7 Bc3 Bf1 Kf2 Ba6 Kg2 Be2 Bd4 Bb5 Bc3 Bd6 Kh2 Bb8 Bd4 Bc7 Kg2 Ba5 Bc3 Bd8 Bd4 Bc7 Kh2 Ba5 Bc3 Bd8 Kg2 Bc6+ Kh2 Bc7 Bd4 Bd6 Nc4 Be7 Ne5 Bf8 Kh3 Bd6 Kh2 Bc7 Bc3 Be8 Bd4 Bd8 Kh3 Be7 Bc;
7r/p3k3/2p5/1pPp4/3P4/PP4P1/3P1PB1/2K5 w - - acd 68; acs 3468; bm Bf3; c0 "I think that Bf3 also draws. 68 plies and the engines can make no progress."; c1 "diff=10"; c3 "Kd1"; c4 "(Adorjan_EofCE3) #934 draw"; ce -205; id "EG 2059"; pm Bf3; pv Bf3 Rf8 Bg4 Rxf2 b4 Rf1+ Kc2 Kd8 Kb3 Kc7 Kc3 Rc1+ Kb3 Rg1 Bf5 Rxg3+ Kb2 Rf3 Bc2 Rf4 Kc3 a6 Bd1 Rf1 Bc2 Rf3+ Kb2 Rf2 Kc3 Kd7 Bb1 Rf6 Kb2 Kc8 Bc2 Rf2 Kc3 Rf3+ Kb2 Kb7 Bg6 Rf4 Kc3 Rh4 Bf5 Rh1 Bg6 Ra1 Kb3 Rf1 Kc3 Kb8 Be8 Rf6 Bh5 Rh6 Bg4 Rg6 Bf5 Rg3+ Kb2 Kb7 d;
7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - acd 44; acs 3228; bm Nb7; c0 "Incredibly difficult. Zugzwang involved."; c1 "diff=10"; c3 "hxg6"; ce 0; id "yacpdb id 340530"; pm Nb7; pv Nb7 Ke7+ Ka2 Qc3 Rd1 Qb4 Bc2 Qxc4+ Kb2 Qb4+ Bb3 bxc5 hxg6 c4 Nxa5 Qxa5 Bxc4 Qxa4 Bb3 Qxb5 Rh7+ Kf6 g7 Qe5+ Kb1 Qb5 Kb2;
7k/pp2Np1p/2p2P1N/5pP1/7K/P4n2/1q3Q2/5n2 w - - acd 66; acs 3286; bm Qxf3; c1 "diff=10"; c3 "Kh5"; ce 0; pm Qxf3; pv Qxf3 Qh2+ Qh3 Qf2+ Kh5 Qe2+ Kh4 Qe1+ Kh5 Qd1+ Kh4 Qe1+;
4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - acd 52; acs 3487; bm Qa3; c0 "level: hard-16"; c1 "diff=10"; c3 "Qxe5"; ce 0; id "MG 4322"; pm Qa3; pv Qa3 Rxe6 Qxa7 Qe7 Qa8+ Qe8 Qb7 Qe7 Qa8+;
4bbk1/3nq2r/pr2p3/Np1p1p1p/1PpP1Pp1/2P1P1P1/PQ1NB1PR/2K4R b - - acd 51; acs 770; bm Nf6; c0 "Alexander Ipatov - Vladimir Kramnik (Turkey - Russia, round 6) WchT 9th 29...e5 TP 2013"; c1 "diff=08"; c3 "e5"; ce 167; pm Nf6; pv Nf6 Bd1 Rb8 Bc2 Qc7 Bd1 Bd6 Kc2 Qd7 Be2 Bg6 Kc1 Be7 Kb1 Rd8 Ka1 Rc8 Rb1 Bd8 Rhh1 Rb8 Rbc1 Bb6 Rb1 Bc7 Rh2 Ne8 Qa3 Bb6 Qb2 Qc7 Qa3 Rc8 Rhh1 Rd8 Rbe1 Nf6 Rb1 Qd7 Qb2 Ne8 Qa3 Rb8 Rbe1 Bd8 Rc1 Bc7 Qb2 Bb6 Rh2 Bc7;
2b1r3/r2ppN2/8/1p1p1k2/pP1P4/2P3R1/PP3PP1/2K5 w - - acd 46; acs 938; bm Nh6+; c1 "diff=10"; c3 "Nd6+"; ce -375; id "Vincent Lejeune - hard CCC 2008 - Pos 51"; pm Nh6+; pv Nh6+ Kf6 Ng8+ Ke6 Nh6 Rf8 Re3+ Kd6 Ng4 Rg8 f3 Ra8 Re2 Rh8 Re1 Bb7 Ne3 e6 Kc2 Rh2 Rf1 Ke7 Kd3 Rh4 Re1 Kd6 Rf1 Rah8 Re1 Rh1 Ke2 R1h7 Kd3 Rh2 Re2 Rh1 Nc2 Rg8 Ke3 Rh2 Ne1 Rc8 Kd2 Ra8 Nd3 Ke7 Nc5 Bc6 g4 Rxe2+ Kxe2 Rh8 Kd3 Rh2;
1R6/8/8/5bp1/4p2k/8/B1p2PKP/8 w - - acd 62; acs 3255; bm Rh8+; c0 "Vitaly Chekhover, Gatchinskaja Pravda, 1954 White to play and draw (is 1.Rh8+ Kg4 2.Rf8 c1Q 3.h3+ Kf4 4.Be6 Ke5 5.Bxf5 Qa3 6.Rf7 Qf3+ 7.Kg1 e3 8.Re7+ Kf6 9.Rxe3 Qxf5 an alternate solution?)"; c1 "diff=10"; c3 "Rb1"; c4 "(Speelman_EP) (an example of building a fortress to get a draw, p38, 72)"; ce -424; id "(Speelman_EP) (an example of building a fortress to get a draw, p38, 72)"; pm Rh8+; pv Rh8+ Kg4 Rf8 c1=Q h3+ Kf4 Be6 Ke5 Bxf5 Qa3 Rf7 Qf3+ Kg1 e3 Re7+ Kf4 Rxe3 Qd1+ Kg2 Kxf5 Rf3+ Ke4 Re3+ Kf4 Rc3 Qd5+ Kg1 Qd2 Rg3 Qb2 Kg2 Qb7+ Kg1 Kf5 Re3 Qd5 Rc3 Kf4 Re3 Qd1+ Kg2 Qc2 Rf3+ Ke5 Re3+ Kf6 Rf3+ Kg7 Re3 Qc6+ Kg1 Kf6 Re1 Qc8 Re3 Qf5 Rg3 Qb1+ Kg2 Qa;
1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - acd 65; acs 3464; bm Ke6; c0 "level: hard-16"; c1 "diff=09"; c3 "Rxb3+"; ce 198; id "Pos.4 Brzozka-Bronstein, Miskolc 1963"; pm Ke6; pv Ke6 Bc2 Rb4 Bb1 Rc6 Bc2 Rb8 Bd1 Ra8 Bc2 Rc7 Bd1 Rb8 Be2 Rcb7 Bd1 Rc8 Be2 Rbb8 Rd1 Rb4 Rd5 Ra8 Bd1 Rb7 Be2 Rh7 Rd1 Rhh8 Rd5 Rhb8 Bd1 Rb4 Be2 Rb7 Bf1 Raa7 Rd1 Rh7 Rd5 Ra8 Bg2 Rhh8 Rd1 Rhb8 Bf1 Ra7 Be2 Rh8 Rd5 Rc7 Bf1 Ra8 Be2 Rcc8 Rd1 Ke7 Rd5 Rh8 Rd1 Rab8 Rd;
1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - acd 127; acs 141; bm Kd7; c3 "Kd5"; ce 0; dm 27; id "Vincent Lejeune - hard CCC 2008 - Pos 29"; pm Kd7; pv Kd7 Qg8 Kc6 Qh8;

[Dann Corbit]

I should mention that the proposed solutions are in field c3.

[Vinvin]

These are from Vincent Lejeune's recent set of 115 chess problems. I have mechanical (computer generated) solutions to all but 11 of them. Some are incredibly difficult.

For instance, with this problem:
http://www.yacpdb.org/?id=340530
I tried creation of all the positions to the solution and then analyzing them backwards to feed the hash table. Eventually, the solution gets so far away that the engines get lost anyway.

If anyone has mechanical solutions to any of these problems, I would be very glad to receive them.

8/4ppkp/3p2p1/2pP1PP1/2P1P1P1/p7/K3B3/2b5 b - - acd 60; acs 1215; bm Bxg5; c0 "Bxg5 definitely wins. Does f6 win faster?"; c1 "diff=08"; c3 "f6"; c4 "(Dvor) 163"; ce 549; id "Dvoretsky.126"; pm Bxg5; pv Bxg5 fxg6 fxg6 Kxa3 Kf6 Kb2 Ke5 Kc2 Kxe4 Kd1 Bh4 Kd2 Kd4 Bf1 Bg3 Be2 Bf2 Bf1 Bh4 Be2 Bg3 Kc2 Ke3 Bd1 Be5 Kb3 Kd3 Bf3 Bf6 Bd1 Kd2 Bf3 Bd4 Ka3 Ke3 Bd1 Kf4 Be2 h5 gxh5 g5 Bf1 Kf3 h6 g4 h7 g3 Kb3 g2 Bxg2+ Kxg2 Kc2 Bf6 Kd3 Kf3 Kd2 Ke4 Kd1 Kd3 Ke1 Kxc4 Ke2 Kxd5;
-> If it's right, I'll take out this one.

8/1p2b3/6k1/5p2/p2NbP1p/P1B1P1p1/1P5P/6K1 w - - acd 65; acs 2720; bm Be1; c0 "Event:Smirin vs Computers; Site:KC-CB INT; Date:2002.04.24; Round: 6; Smirin, Ilia vs Hiarcs 8"; c1 "diff=10"; c3 "h3"; ce -26; pm Be1; pv Be1 gxh2+ Kxh2 Kh5 Kh3 Bd3 Nf3 Bf1+ Kh2 Bc5 Bd2 Bb6 Ne5 Bb5 Kg2 Bc5 Kf3 Be7 Bc3 Bf1 Kf2 Ba6 Kg2 Be2 Bd4 Bb5 Bc3 Bd6 Kh2 Bb8 Bd4 Bc7 Kg2 Ba5 Bc3 Bd8 Bd4 Bc7 Kh2 Ba5 Bc3 Bd8 Kg2 Bc6+ Kh2 Bc7 Bd4 Bd6 Nc4 Be7 Ne5 Bf8 Kh3 Bd6 Kh2 Bc7 Bc3 Be8 Bd4 Bd8 Kh3 Be7 Bc;

7r/p3k3/2p5/1pPp4/3P4/PP4P1/3P1PB1/2K5 w - - acd 68; acs 3468; bm Bf3; c0 "I think that Bf3 also draws. 68 plies and the engines can make no progress."; c1 "diff=10"; c3 "Kd1"; c4 "(Adorjan_EofCE3) #934 draw"; ce -205; id "EG 2059"; pm Bf3; pv Bf3 Rf8 Bg4 Rxf2 b4 Rf1+ Kc2 Kd8 Kb3 Kc7 Kc3 Rc1+ Kb3 Rg1 Bf5 Rxg3+ Kb2 Rf3 Bc2 Rf4 Kc3 a6 Bd1 Rf1 Bc2 Rf3+ Kb2 Rf2 Kc3 Kd7 Bb1 Rf6 Kb2 Kc8 Bc2 Rf2 Kc3 Rf3+ Kb2 Kb7 Bg6 Rf4 Kc3 Rh4 Bf5 Rh1 Bg6 Ra1 Kb3 Rf1 Kc3 Kb8 Be8 Rf6 Bh5 Rh6 Bg4 Rg6 Bf5 Rg3+ Kb2 Kb7 d;
->Nice ! I had no doubt about this one !

7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - acd 44; acs 3228; bm Nb7; c0 "Incredibly difficult. Zugzwang involved."; c1 "diff=10"; c3 "hxg6"; ce 0; id "yacpdb id 340530"; pm Nb7; pv Nb7 Ke7+ Ka2 Qc3 Rd1 Qb4 Bc2 Qxc4+ Kb2 Qb4+ Bb3 bxc5 hxg6 c4 Nxa5 Qxa5 Bxc4 Qxa4 Bb3 Qxb5 Rh7+ Kf6 g7 Qe5+ Kb1 Qb5 Kb2;
-> Very very very deep, may be feasible for engines in 10 years ...

7k/pp2Np1p/2p2P1N/5pP1/7K/P4n2/1q3Q2/5n2 w - - acd 66; acs 3286; bm Qxf3; c1 "diff=10"; c3 "Kh5"; ce 0; pm Qxf3; pv Qxf3 Qh2+ Qh3 Qf2+ Kh5 Qe2+ Kh4 Qe1+ Kh5 Qd1+ Kh4 Qe1+;
-> a lot of zugzwang's (may be with "nullmoves=off" to check there's some hope for engines ?)

4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - acd 52; acs 3487; bm Qa3; c0 "level: hard-16"; c1 "diff=10"; c3 "Qxe5"; ce 0; id "MG 4322"; pm Qa3; pv Qa3 Rxe6 Qxa7 Qe7 Qa8+ Qe8 Qb7 Qe7 Qa8+;
-> here some engines find the move but with a draw score. The point is to find a winning score. I'll sort positions who need to check for positive score.

4bbk1/3nq2r/pr2p3/Np1p1p1p/1PpP1Pp1/2P1P1P1/PQ1NB1PR/2K4R b - - acd 51; acs 770; bm Nf6; c0 "Alexander Ipatov - Vladimir Kramnik (Turkey - Russia, round 6) WchT 9th 29...e5 TP 2013"; c1 "diff=08"; c3 "e5"; ce 167; pm Nf6; pv Nf6 Bd1 Rb8 Bc2 Qc7 Bd1 Bd6 Kc2 Qd7 Be2 Bg6 Kc1 Be7 Kb1 Rd8 Ka1 Rc8 Rb1 Bd8 Rhh1 Rb8 Rbc1 Bb6 Rb1 Bc7 Rh2 Ne8 Qa3 Bb6 Qb2 Qc7 Qa3 Rc8 Rhh1 Rd8 Rbe1 Nf6 Rb1 Qd7 Qb2 Ne8 Qa3 Rb8 Rbe1 Bd8 Rc1 Bc7 Qb2 Bb6 Rh2 Bc7;

2b1r3/r2ppN2/8/1p1p1k2/pP1P4/2P3R1/PP3PP1/2K5 w - - acd 46; acs 938; bm Nh6+; c1 "diff=10"; c3 "Nd6+"; ce -375; id "Vincent Lejeune - hard CCC 2008 - Pos 51"; pm Nh6+; pv Nh6+ Kf6 Ng8+ Ke6 Nh6 Rf8 Re3+ Kd6 Ng4 Rg8 f3 Ra8 Re2 Rh8 Re1 Bb7 Ne3 e6 Kc2 Rh2 Rf1 Ke7 Kd3 Rh4 Re1 Kd6 Rf1 Rah8 Re1 Rh1 Ke2 R1h7 Kd3 Rh2 Re2 Rh1 Nc2 Rg8 Ke3 Rh2 Ne1 Rc8 Kd2 Ra8 Nd3 Ke7 Nc5 Bc6 g4 Rxe2+ Kxe2 Rh8 Kd3 Rh2;
-> fortress, it's a job for Sting SF in some months.

1R6/8/8/5bp1/4p2k/8/B1p2PKP/8 w - - acd 62; acs 3255; bm Rh8+; c0 "Vitaly Chekhover, Gatchinskaja Pravda, 1954 White to play and draw (is 1.Rh8+ Kg4 2.Rf8 c1Q 3.h3+ Kf4 4.Be6 Ke5 5.Bxf5 Qa3 6.Rf7 Qf3+ 7.Kg1 e3 8.Re7+ Kf6 9.Rxe3 Qxf5 an alternate solution?)"; c1 "diff=10"; c3 "Rb1"; c4 "(Speelman_EP) (an example of building a fortress to get a draw, p38, 72)"; ce -424; id "(Speelman_EP) (an example of building a fortress to get a draw, p38, 72)"; pm Rh8+; pv Rh8+ Kg4 Rf8 c1=Q h3+ Kf4 Be6 Ke5 Bxf5 Qa3 Rf7 Qf3+ Kg1 e3 Re7+ Kf4 Rxe3 Qd1+ Kg2 Kxf5 Rf3+ Ke4 Re3+ Kf4 Rc3 Qd5+ Kg1 Qd2 Rg3 Qb2 Kg2 Qb7+ Kg1 Kf5 Re3 Qd5 Rc3 Kf4 Re3 Qd1+ Kg2 Qc2 Rf3+ Ke5 Re3+ Kf6 Rf3+ Kg7 Re3 Qc6+ Kg1 Kf6 Re1 Qc8 Re3 Qf5 Rg3 Qb1+ Kg2 Qa;
Nice little fortress after a couple of moves 1...e2?! is only draw ...
[d]8/8/8/5Bp1/7k/4p3/5PKP/8 b - - 0 1

1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - acd 65; acs 3464; bm Ke6; c0 "level: hard-16"; c1 "diff=09"; c3 "Rxb3+"; ce 198; id "Pos.4 Brzozka-Bronstein, Miskolc 1963"; pm Ke6; pv Ke6 Bc2 Rb4 Bb1 Rc6 Bc2 Rb8 Bd1 Ra8 Bc2 Rc7 Bd1 Rb8 Be2 Rcb7 Bd1 Rc8 Be2 Rbb8 Rd1 Rb4 Rd5 Ra8 Bd1 Rb7 Be2 Rh7 Rd1 Rhh8 Rd5 Rhb8 Bd1 Rb4 Be2 Rb7 Bf1 Raa7 Rd1 Rh7 Rd5 Ra8 Bg2 Rhh8 Rd1 Rhb8 Bf1 Ra7 Be2 Rh8 Rd5 Rc7 Bf1 Ra8 Be2 Rcc8 Rd1 Ke7 Rd5 Rh8 Rd1 Rab8 Rd;
-> Engines have to see that the position is draw (fortress) with the 2 black rooks and then see the endgame is winning after the sac Rxb3!!

1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - acd 127; acs 141; bm Kd7; c3 "Kd5"; ce 0; dm 27; id "Vincent Lejeune - hard CCC 2008 - Pos 29"; pm Kd7; pv Kd7 Qg8 Kc6 Qh8;
-> even after 1. Kd5 Kc7 2.Bb5, Stockfish have trouble to see black is lost after Be8-Bf7.
[d]5b1q/pPk1p1p1/P3P1Pp/1B1K3P/8/8/8/8 b - - 3 2

[zullil]

If anyone has mechanical solutions to any of these problems, I would be very glad to receive them.

[d]1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - -

acd 127; acs 141; bm Kd7; c3 "Kd5"; ce 0; dm 27; id "Vincent Lejeune - hard CCC 2008 - Pos 29"; pm Kd7; pv Kd7 Qg8 Kc6 Qh8;

Not sure what you want. Have you tried using Sting or Stockfish Matefinder? Or simply disable nullmove pruning in Stockfish and recompile to get

Code: Select all

info depth 54 seldepth 64 multipv 1 score mate 27 nodes 19832454462 nps 35275883 hashfull 999 tbhits 0 time 562210 pv c4b5 h8g8 c6d5 b8c7 b5e8 c7b8 e8f7 g8h8 d5c6 h8g8 c6d7 g8h8 d7e8 b8c7 b7b8q c7b8 e8d7 b8a8 d7c7 h8g8 c7c8 g8f7 g6f7 g7g5 h5g6 h6h5 c8d7 a8b8 d7e8 f8g7 e8e7 b8c7 f7f8q g7f8 e7f8 h5h4 e6e7 h4h3 e7e8q h3h2 e8e5 c7c6 e5h2 c6b5 g6g7 b5b4 h2d2 b4b5 g7g8q b5a6 d2b2 a6a5 g8a2

Not sure this is minimal, since I haven't studied the output at all.

[Ferdy]

These are from Vincent Lejeune's recent set of 115 chess problems. I have mechanical (computer generated) solutions to all but 11 of them.

What about using aquarium IDEA? less Interactive, by just setting the position analyze it with engine then just leave it for some hours,
after which just follow the minimized move scores, to create the pv.

I tried creation of all the positions to the solution and then analyzing them backwards to feed the hash table.

How did you do this? Did you use some GUI's say arena?

There are pv's in the epd lines are those output generated by the engine
when analyzing the root position?

[peter]

For instance, with this problem:
http://www.yacpdb.org/?id=340530
I tried creation of all the positions to the solution and then analyzing them backwards to feed the hash table.

Hi Dann!

Mihai Neghina first showed me that great problem of his, you gave the link to, back in 2010.

[d]7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - 0 1

I don't think an engine with default- params can even keep the way to solution backward in hash nowadays.
Not even, as Louis suggested too, with switched off nullmove and backward solving, and then that's very much human intervention to "help" the engine, isn't it?
Did I get you right, that all of this kind was "allowed" as for the "mechanical solutions"?
If you give some positions only to fill the hash but don't give lines to solve backward, it yet could be all the positions along the lines of solution, couldn't it?


E.g. in Neghinas study of your given link and the diagram above a SF with switched off nullmove, very little futility pruning and also lower LMR didn't manage to keep a winning score in output further back then to move 9 after backward solving of the solution line with 32G hash, because at the latest there the engine gets lost again in a side line.

Thanks for sharing

[BBauer]

It may not be a good idea to disable null move completely.
Instead I do the following.
Read in the variable LegalMov_GT as an UCI option.
Then you may set another variable MobilityVal by

if (LegalMov_GT == 0) MobilityVal=0;
else MobilityVal=MoveList<LEGAL>(pos).size();

before Step 8 in search.cpp.
And add
&& MobilityVal >= LegalMov_GT
to the if statement in Step 8
If you set LegalMov_GT to 0 nothing changes, but if you set
LegalMov_GT to say 11 you may get:
[d]1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1

Code: Select all

St-mod-15-04-22:
 38/5	00:00	 929k	1.397k	 0,00	Kc6-d7 Dh8-g8 Kd7-c6 Dg8-h8
 39/46	00:03	 6.693k	1.956k	+8,67	Kc6-d5 Dh8-g8 Kd5-e5 Dg8-h8 Lc4-b5 Dh8-g8
 40/49+	00:03	 6.859k	1.935k	+8,73	Kc6-d5
 40/49-	00:03	 7.464k	1.869k	+8,67	Kc6-d5 Dh8-g8
 40/49-	00:04	 8.949k	1.903k	+7,89	Kc6-d5 Dh8-g8
 40/52+	00:05	 10.775k	1.878k	+8,13	Kc6-d5
 40/56+	00:13	 30.976k	2.242k	+30,62	Kc6-d5
 40/58	00:17	 39.996k	2.293k	+31,81	Kc6-d5 Dh8-g8 Lc4-b5 Kb8-c7 Lb5-e8
 41/58-	00:17	 41.057k	2.301k	+31,75	Kc6-d5 Dh8-g8
 41/58+	00:17	 41.409k	2.302k	+31,81	Kc6-d5
 41/58-	00:18	 43.165k	2.310k	+31,75	Kc6-d5 Dh8-g8
 41/58+	00:19	 44.101k	2.316k	+31,83	Kc6-d5
 41/58+	00:22	 52.033k	2.300k	+42,05	Kc6-d5
 41/58-	00:22	 52.877k	2.300k	+39,02	Kc6-d5 Dh8-g8
 41/58-	00:23	 53.481k	2.310k	+31,00	Kc6-d5 Dh8-g8
 41/58+	00:23	 55.255k	2.325k	+37,22	Kc6-d5
 41/58+	00:26	 61.571k	2.361k	+55,12	Kc6-d5
 41/58	00:27	 64.967k	2.367k	+55,12	Kc6-d5 Dh8-g8 Lc4-b5 Kb8-c7 Lb5-a4

Kind regards
Bernhard

[lech]

For instance, with this problem:
http://www.yacpdb.org/?id=340530
I tried creation of all the positions to the solution and then analyzing them backwards to feed the hash table.

Hi Dann!

Mihai Neghina first showed me that great problem of his, you gave the link to, back in 2010.

[d]7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - 0 1

I don't think an engine with default- params can even keep the way to solution backward in hash nowadays.
Not even, as Louis suggested too, with switched off nullmove and backward solving, and then that's very much human intervention to "help" the engine, isn't it?
Did I get you right, that all of this kind was "allowed" as for the "mechanical solutions"?
If you give some positions only to fill the hash but don't give lines to solve backward, it yet could be all the positions along the lines of solution, couldn't it?


E.g. in Neghinas study of your given link and the diagram above a SF with switched off nullmove, very little futility pruning and also lower LMR didn't manage to keep a winning score in output further back then to move 9 after backward solving of the solution line with 32G hash, because at the latest there the engine gets lost again in a side line.

Thanks for sharing

A part of problems of this study I described here: http://talkchess.com/forum/viewtopic.php?t=56045
(after: 1. hxg6 Qxh2 2. Rxe6+ Kxe6 3. g7 Qh1+ 4. Kb2 Qa8 5. Bxf5+ Kf6
6. Sc8 Kxg7 7. c6 Kf6 )

Here are two problems:
A sacrifice leading to fortress and a walking king.

Maybe ..., the next Sting will solve this position, of course without mechanical tricks.

[peter]

[Here are two problems:
A sacrifice leading to fortress and a walking king.

Maybe ..., the next Sting will solve this position, of course without tricks.

Of course this study is a typical anti- engine- problem, a thing in which Mihai Neghina is one of the very great masters to me.

To the weaknesses of the engines you describe as for this position, nullmove is to be counted in additionally for sure too.

If your next Sting indeed would solve such from the starting position on its own, that would be a really great progress, I'd say.
Go, Marek, go!

Till now, Sting5 isn't even able to solve it with going backward from move nr.20, when the winning score is found easily, furhter back then to move nr.10 on good hardware, at latest there the score drops to 0.00 again and again, losing itself in side lines, even if trying those several times and moves forward-backward again and again

[lech]

[Here are two problems:
A sacrifice leading to fortress and a walking king.

Maybe ..., the next Sting will solve this position, of course without tricks.

Of course this study is a typical anti- engine- problem, a thing in which Mihai Neghina is one of the very great masters to me.

To the weaknesses of the engines you describe as for this position, nullmove is to be counted in additionally for sure too.

If your next Sting indeed would solve such from the starting position on its own, that would be a really great progress, I'd say.
Go, Marek, go!

Till now, Sting5 isn't even able to solve it with going backward from move nr.20, when the winning score is found easily, furhter back then to move nr.10 on good hardware, at latest there the score drops to 0.00 again and again, losing itself in side lines, even if trying those several times and moves forward-backward again and again

Without a special code for walking Kings it is impossible to solve this study.
The next Sting will cinttain such a "wise" code and it can be solved.