These positions are from various test sets by Walter Eigenmann. How many of them can your favorite engine solve?
Stockfish has an awful time with these.
r1n2N1k/2n2K1p/3pp3/5Pp1/b5R1/8/1PPP4/8 w - - bm Ng6+;
4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - bm Bg7;
8/2pN1k2/p4p1p/Pn1R4/3b4/6Pp/1P3K1P/8 w - - bm Ke1;
8/8/4p1Pk/1rp1K1p1/4P1P1/1nP2Q2/p2b1P2/8 w - - bm Kf6;
2kb4/p7/r1p3p1/p1P2pBp/R2P3P/2K3P1/5P2/8 w - - bm Bxd8;
r1k1n2n/8/pP6/5R2/8/1b1B4/4N3/1K5N w - - bm b7;
3r2k1/p1R2ppp/1p6/P1b1PP2/3p4/3R2B1/5PKP/1r6 w - - bm f6;
r1br2k1/7p/p2q1ppB/1p1P4/2P5/5N2/PQ3PPP/4R1K1 w - - bm Ng5;
4r1k1/pp5p/2p1bR2/2P5/2B1n3/2B2K2/PP5P/8 w - - bm Rxe6;
1k5r/pb3p1p/1q2rp1b/2p5/1pQ1P3/1P3N2/P3BPPP/3R1RK1 w - - bm Rd5;
2r1rbk1/p1qn2p1/3n1pp1/P1p1p3/1p2P1P1/1N3Q1P/1PP2P2/R1BR1BK1 w - - bm Rxd6;
r1r3k1/1bq2pp1/p4n1p/1p1npB2/PP6/2P2N2/1BQ2PPP/R3R1K1 b - - bm Nf4;
3rrnk1/p1n1qpp1/1p5p/2pRP3/2P4P/1P2Q1PB/PB5K/4R3 w - - bm Red1;
r1b2rk1/pp3pb1/3q2pp/2nP4/6P1/2P2N2/P1Q1BPP1/2BR1RK1 w - - bm Rd4;
5rk1/ppqb2pp/2n1p3/3pP3/2pP4/P1P1BN1Q/2P2PP1/R3K3 b Q - bm Rxf3;
2k5/1bp2B2/1b5r/1N1Pp2p/6r1/6B1/1P2RP2/R5K1 b - - bm h4;
r4rk1/ppnqp2p/3p2p1/2pP4/P1P2B1P/6P1/1P3PK1/R2Q1R2 b - - bm Rxf4;
r1q2k2/pppb3r/2n1pBp1/3pPp1p/3P1Q1P/P2B4/2P2PP1/R3K2R w KQ - bm g4;
r3kb1r/4Pp2/pn3P2/1pp3B1/2p5/2N3P1/PP3P1P/3RKB1b w kq - bm h4;
1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - bm Rxb3+;
r3rnk1/1bq1bpp1/pp2p3/3pP2Q/P4p2/1NNB1R2/1PP4P/2KR4 w - - bm Rg1;
4K2k/1pQp3p/1R4p1/1P4P1/8/3P3P/1q3r2/8 w - - bm Qd8;
5B2/1p1nNp2/8/1P2R3/3p3q/8/p1K1P1Q1/k3b3 w - - bm Ng6;
2r1k1r1/3q4/p1nBpp1p/8/B5p1/4RbK1/PP1Q1P1P/2R5 w - - bm h4;
6r1/6r1/2p1k1pp/p1pbP2q/Pp1p1PpP/1P1P2NR/1KPQ3R/8 b - - bm Qf5;
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
seems like a set that would need 5-10mins/position to see much of significance in the way of results, even h5 tactical is flailing quite badly at 1min/position
yanquis1972 wrote:seems like a set that would need 5-10mins/position to see much of significance in the way of results, even h5 tactical is flailing quite badly at 1min/position
I am running Houdini and Komodo at about 1/2 hour per position on 10 cores.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Thanks for posting. I find it fascinating to see the evidence of how much room for improvement remains, despite an already phenomenal software progress.
Another thought is the similarity of the leading programs today. You would think that various different positions would succumb to different programs, but it appears that they all have difficulties with this suite.
Here are the EPD records with detail added from the analysis.
The last record for the Komodo task failed, because the chess tasks get killed on that machine at 2AM and I miscalculated how much time to allot.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
r4rk1/ppnqp2p/3p2p1/2pP4/P1P2B1P/6P1/1P3PK1/R2Q1R2 b - - bm Rxf4;
r3rnk1/1bq1bpp1/pp2p3/3pP2Q/P4p2/1NNB1R2/1PP4P/2KR4 w - - bm Rg1;
r3kb1r/4Pp2/pn3P2/1pp3B1/2p5/2N3P1/PP3P1P/3RKB1b w kq - bm h4;
r1r3k1/1bq2pp1/p4n1p/1p1npB2/PP6/2P2N2/1BQ2PPP/R3R1K1 b - - bm Nf4;
r1br2k1/7p/p2q1ppB/1p1P4/2P5/5N2/PQ3PPP/4R1K1 w - - bm Ng5;
r1b2rk1/pp3pb1/3q2pp/2nP4/6P1/2P2N2/P1Q1BPP1/2BR1RK1 w - - bm Rd4;
5rk1/ppqb2pp/2n1p3/3pP3/2pP4/P1P1BN1Q/2P2PP1/R3K3 b Q - bm Rxf3;
4r1k1/pp5p/2p1bR2/2P5/2B1n3/2B2K2/PP5P/8 w - - bm Rxe6;
4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - bm Bg7;
4K2k/1pQp3p/1R4p1/1P4P1/8/3P3P/1q3r2/8 w - - bm Qd8;
3rrnk1/p1n1qpp1/1p5p/2pRP3/2P4P/1P2Q1PB/PB5K/4R3 w - - bm Red1;
3r2k1/p1R2ppp/1p6/P1b1PP2/3p4/3R2B1/5PKP/1r6 w - - bm f6;
2r1rbk1/p1qn2p1/3n1pp1/P1p1p3/1p2P1P1/1N3Q1P/1PP2P2/R1BR1BK1 w - - bm Rxd6;
2r1k1r1/3q4/p1nBpp1p/8/B5p1/4RbK1/PP1Q1P1P/2R5 w - - bm h4;
2k5/1bp2B2/1b5r/1N1Pp2p/6r1/6B1/1P2RP2/R5K1 b - - bm h4;
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Sample results from my i7-2600K, engines using single core. I am surprised with CDrill's perf, since this engine has no transposition table, no null move pruning, no LMR rated around 1800 on CCRL 40/4.
Yes, I have the eret suite. The only one unsolved for me is:
[d]4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - acd 34; acs 1214; bm Be3; c3 "Bg7"; ce 38; id "ERET 084 - Koenigsangriff"; pm Be3; pv Be3 Nf8 Bxc4+;
For all the others, I have a mechanical solution that agrees with the proposed solution.
I have 383 positions from various test suites that he wrote and I have a total of 15 where I did not come up with the same answer as the proposed answer which also gives the best score.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.