Rybka or... Franz Huber's Chest UCI?

[swami]

I remember Franz Huber's checkmate solving program called Chest few years ago, it was the only program that can solve long range checkmates that other engines fail to see.

I wonder if Rybka 3 can see long range mates better than Mr.Huber's Chest UCI in a difficult checkmate problems?

I'd like to see someone doing an experiment on this.

[M ANSARI]

Well it all depends. Chest is exclusively designed to look for a mate and nothing else. Rybka is designed to look for the strongest move. I think Rybka Winfinder works similar to Chest. Also you could probably setup Rybka 3 with Monte Carlo to solve even the most complicated mating position.

[Dann Corbit]

I remember Franz Huber's checkmate solving program called Chest few years ago, it was the only program that can solve long range checkmates that other engines fail to see.

I wonder if Rybka 3 can see long range mates better than Mr.Huber's Chest UCI in a difficult checkmate problems?

I'd like to see someone doing an experiment on this.

They are answering different questions.

Chest is a mate proof system. When it finds a solution you are guaranteed that there are no better solutions.

When a chess engine finds a solution, you will never know if there is a better one or not.

[swami]

I remember Franz Huber's checkmate solving program called Chest few years ago, it was the only program that can solve long range checkmates that other engines fail to see.

I wonder if Rybka 3 can see long range mates better than Mr.Huber's Chest UCI in a difficult checkmate problems?

I'd like to see someone doing an experiment on this.

They are answering different questions.

Chest is a mate proof system. When it finds a solution you are guaranteed that there are no better solutions.

When a chess engine finds a solution, you will never know if there is a better one or not.

So by your reasoning, I'd guess Chest UCI is the better and quicker checkmate solver than Rybka windfinder if they were given the difficult checkmate problems, and have been set to take as much time they wanted to.

Are there any long range checkmate problems available somewhere? Especially the ones that are difficult for engines like Rybka? I don't know. Perhaps you know better.

I saw an interesting case in CCT11, Bright at one point was seeing checkmate in 10 against Gaviota, but the latter was just evaluating the scores several moves ahead without the foreknowledge of any checkmate coming nearer.

[swami]

Well it all depends. Chest is exclusively designed to look for a mate and nothing else. Rybka is designed to look for the strongest move. I think Rybka Winfinder works similar to Chest. Also you could probably setup Rybka 3 with Monte Carlo to solve even the most complicated mating position.

Yeah It would make an interesting experiment.. but the problem is that nobody has any such test suites of long range checkmates. Perhaps there have some problems posted into CCC, but never enough to collect the set of challenging problems.

I sometimes wonder what the criteria is to select checkmate puzzles, since most of the games end via tablebases anyway. Perhaps I should go look into games where quick wins happen in middlegame. But that too have many cases of engine resigning in advance.

[Marek Soszynski]

As soon as an ordinary engine finds a single forced mate, it is "job done" - some engines don't even bother analysing any further. But because of the way engines search, the first mate found might not be the shortest. A dedicated mate solver has to look at everything (so is very slow in comparison) but will find the genuinely shortest mate, and also find out whether it is a unique solution. I smile when I see posters boasting that one engine or another (without tablebases) sees a mate in 24, say. Mate announcements by ordinary engines are not to be entirely trusted.

[Dann Corbit]

I remember Franz Huber's checkmate solving program called Chest few years ago, it was the only program that can solve long range checkmates that other engines fail to see.

I wonder if Rybka 3 can see long range mates better than Mr.Huber's Chest UCI in a difficult checkmate problems?

I'd like to see someone doing an experiment on this.

They are answering different questions.

Chest is a mate proof system. When it finds a solution you are guaranteed that there are no better solutions.

When a chess engine finds a solution, you will never know if there is a better one or not.

So by your reasoning, I'd guess Chest UCI is the better and quicker checkmate solver than Rybka windfinder if they were given the difficult checkmate problems, and have been set to take as much time they wanted to.

Are there any long range checkmate problems available somewhere? Especially the ones that are difficult for engines like Rybka? I don't know. Perhaps you know better.

I saw an interesting case in CCT11, Bright at one point was seeing checkmate in 10 against Gaviota, but the latter was just evaluating the scores several moves ahead without the foreknowledge of any checkmate coming nearer.

Sometimes Chest will be faster. Sometimes Rybka will be faster.
Also, for EGTB mates, any chess engine that uses them will produce fast and perfect answers (unless it has bugs).

[Dann Corbit]

The Chest problem set is probably the most interesting checkmate problem set.

As far as mate database files go, there are some EPD mate lists here:
http://cap.connx.com/EPD/

For instance, all of these deal with checkmates:
BWTC-3600.epd.bz2
BWTC.EPD.bz2
ENDGMATE.EPD.bz2
Les_Fernandez_Mega_Mates_of_Death.epd.bz2
M20.EPD.bz2
MATEIN2.EPD.bz2
MATESRCH.EPD.bz2
MCI_matesm.epd.bz2
dm.epd.bz2
dm001.epd.bz2
dm002.epd.bz2
dm003.epd.bz2
dm004.epd.bz2
dm005.epd.bz2
dm006.epd.bz2
dm007.epd.bz2
dm008.epd.bz2
dm009.epd.bz2
dm009.out
dm009.rpt
dm010.epd.bz2
dm011.epd.bz2
dm012.epd.bz2
dm013.epd.bz2
dm014.epd.bz2
dm015.epd.bz2
dm016.epd.bz2
dm017.epd.bz2
dm018.epd.bz2
dm019.epd.bz2
dm020.epd.bz2
dm021.epd.bz2
dm022.epd.bz2
dm023.epd.bz2
dm024.epd.bz2
dm025.epd.bz2
dm026.epd.bz2
dm027.epd.bz2
dm028.epd.bz2
dm029.epd.bz2
dm030.epd.bz2
dm031.epd.bz2
dm032.epd.bz2
dm033.epd.bz2
dm034.epd.bz2
dm035.epd.bz2
dm036.epd.bz2
dm037.epd.bz2
dm038.epd.bz2
dm039.epd.bz2
dm040.epd.bz2
dm041.epd.bz2
dm042.epd.bz2
dm043.epd.bz2
dm044.epd.bz2
dm045.epd.bz2
dm046.epd.bz2
dm047.epd.bz2
dm048.epd.bz2
dm050.epd.bz2
dm051.epd.bz2
dm052.epd.bz2
dm053.epd.bz2
dm054.epd.bz2
dm055.epd.bz2
dm056.epd.bz2
dm057.epd.bz2
dm058.epd.bz2
dm060.epd.bz2
dm061.epd.bz2
dm062.epd.bz2
dm063.epd.bz2
dm064.epd.bz2
dm065.epd.bz2
dm066.epd.bz2
dm067.epd.bz2
dm069.epd.bz2
dm070.epd.bz2
dm071.epd.bz2
dm072.epd.bz2
dm074.epd.bz2
dm075.epd.bz2
dm077.epd.bz2
dm082.epd.bz2
dm087.epd.bz2
dm089.epd.bz2
dm092.epd.bz2
dm093.epd.bz2
dm096.epd.bz2
dm1.epd.bz2
dm100.epd.bz2
dm101.epd.bz2
dm102.epd.bz2
dm103.epd.bz2
dm104.epd.bz2
dm105.epd.bz2
dm110.epd.bz2
dm119.epd.bz2
dm120.epd.bz2
dm121.epd.bz2
dm125.epd.bz2
dm126.epd.bz2
dm130.epd.bz2
dm135.epd.bz2
dm14.epd.bz2
dm2.epd.bz2
dm255.epd.bz2
dm3.epd.bz2
dm4.epd.bz2
dm5.epd.bz2
dm9.epd.bz2
easy-mates.epd.bz2
easy-mates.out.bz2
emates.epd.bz2
m1.epd.bz2
m10.epd.bz2
m10a.epd.bz2
m11.epd.bz2
m12.epd.bz2
m15.epd.bz2
m16.epd.bz2
m2.epd.bz2
m2t.epd.bz2
m3.epd.bz2
m30.epd.bz2
m3a.epd.bz2
m3t.epd.bz2
m7.epd.bz2
m8.epd.bz2
mate.epd.bz2
mates.epd.bz2
matesm.epd.bz2
matetest2.epd.bz2
tmat.epd.bz2
tmate2.epd.bz2
tmates.epd.bz2

[Paloma]

Yeah It would make an interesting experiment.. but the problem is that nobody has any such test suites of long range checkmates. Perhaps there have some problems posted into CCC, but never enough to collect the set of challenging problems.

Hi swami,

here is a database (*.epd) with lots of mates in 2 up to mates in more than 100

http://www.drb.insel.de/~heiner/Chess/others/
or
http://www.geocities.com/fhub55/chess/ChestUCI_V51.zip

kind regards

[Jouni]

You are kidding Swaminathan are You!? Rybka is very weak in mate solving. Here's my old 40 position test (difficult):

b5nq/K2Npp2/2pp1Ppr/2pk4/Q1R2pB1/2P1b3/R2p4/n2r4 w - - bm #6;
6nr/3p1p1r/4P2b/1K1P3n/N1N1P3/1pBk1P1p/bR4P1/8 w - - bm #6;
2r2rk1/2q2p1p/p1p1nPpQ/3p4/4p1B1/8/PPP3PP/R4R1K w - - bm #6;
7n/2r4p/4RPp1/ppkpN3/brp1p3/B1N1P2B/3P3R/7K w - - bm #6;
q4rbr/2b3N1/ppR2p1p/Nn2kp1B/4p2n/PpP1R2P/1PQ3P1/2B3K1 w - - bm #6;
1q2kr2/rB2bbP1/3p1n2/1NpP2BN/P4P1p/1P2Q3/6PP/R3K2R w - - bm #6;
1N1K3Q/3ppPBp/3PP1p1/k1P4r/P1R3P1/3P4/p5bp/1n1Bb1nq w - - bm #6;
r1b2bn1/pNp1pr2/3p2p1/1B1Pk2p/4P3/2N5/PPP2PPP/R1BQK2R w KQ - bm #6;
r3kBbK/brnp4/Bn5Q/6pR/8/8/2R5/8 w q - bm #6;
5R2/2P2pK1/2P2P2/1Pp1BP1P/b6p/1p1RPB2/1p2NPn1/6rk w - - bm #6;
1b5R/2pp2rp/2nP2p1/n1N2pPk/rP2p2P/PR1P3K/B1P5/2B5 w - - bm #7;
k1n3q1/pr1P3p/pP2p1p1/P5p1/1P1P3n/3pp3/K5RB/3Q1R1B w - - bm #7;
4Q3/7K/8/3Pp3/1B2P3/6Pp/6pp/3B1Rbk w - - bm #7;
4r1k1/pp1b1ppp/3b1q2/8/8/1PPp3P/P4P2/RN3KNR b - - bm #7;
2rqrb1n/pb1Rp3/1p4pP/Q1ppk1B1/1n1p4/3P1RKB/P4N1P/8 w - - bm #7;
2b2B2/p4Rp1/2pR2P1/8/7P/1pNP3P/1P1Pp1rp/2K1k2N w - - bm #7;
1N5b/1K6/4B1p1/3N1prb/4P1k1/3R4/3Pp3/4B3 w - - bm #7;
nR2Q1b1/3p1N2/1ppB1pp1/PP3pp1/8/3k4/1Pp5/2K5 w - - bm #7;
8/7b/3R4/2NPN3/PpPP1R2/1Pp1Pp2/p4P1B/k1K5 w - - bm #7;
5R2/2ppB1p1/8/5pNp/5Nb1/3p3p/3P1P1k/R3K3 w Q - bm #7;
3r1r1k/1bq1bp1p/pn5B/1ppPPR2/P1p1B3/2P3P1/7P/R2Q2K1 w - - bm #8;
3b1n2/B3p3/2K1p1p1/4Pkpb/5pn1/2P2N1B/4PP2/6R1 w - - bm #8;
1b6/3p1Bnp/1p1p3k/3p3p/3Pp2P/1N2Np2/K4R2/8 w - - bm #8;
1R1nN3/K1pP1nr1/B1k2Np1/3b4/P4P2/ppPpp1P1/2R5/6r1 w - - bm #8;
r1b3kr/p2pn2p/p2N3B/2QP4/3P4/8/P1PK2PP/q5NR w - - bm #8;
3Q4/2p5/n3B3/2p2P2/PpN1P1p1/1P1Pp3/2Ppp3/bK1krn2 w - - bm #8;
8/8/1p1p1BNP/1p1p3P/1kqp1QRK/3p3P/3p1NBP/8 w - - bm #8;
r1b3k1/pp3r2/1qp1pP1Q/2b5/8/2P5/PPB1RPPP/R5K1 w - - bm #8;
5rk1/pprbbp2/2np3Q/4p1P1/4P3/2qBB3/P1P3P1/3R1R1K w - - bm #8;
8/PPPPPP2/7K/4rP2/8/6Pp/4Nprp/5Qnk w - - bm #8;
1Q6/1p1p4/1p1P2P1/5P1r/1PBbP3/PK1P2pp/2P1p3/knR1B3 w - - bm #9;
2r1k3/6pr/p1nBP3/1p3p1p/2q5/2P5/P1R4P/K2Q2R1 w - - bm #9;
3r3k/pp1bqB1p/2n2R2/4p3/3P3Q/P6P/1BP3P1/1K1R4 w - - bm #9;
6nr/p2rkp1n/1pQRq1pp/1N2P3/8/4BN1P/PPP2PP1/2K4R w - - bm #9;
n2r2k1/P2P1pp1/5r2/2B5/1P6/6Qp/5P1P/R4K2 w - - bm #9;
r1b2r1k/1pq2ppp/p2b1n1N/8/2B1p3/1P3Q2/PBP2PPP/R4R1K w - - bm #9;
r1bq1k1r/1p1pn1bp/3Q2p1/p5p1/2B5/4R3/PP3PPP/R5K1 w - - bm #9;
r4k2/pp1q1p1p/5r1N/2pp2Q1/8/1P1PP3/P2P2P1/5RK1 w - - bm #9;
rn3k1r/p3n1pp/2p2p2/4N3/2Np4/8/q4PPP/1R1QR1K1 w - - bm #9;
rnbq1r1k/pppp1p1p/3n1Pp1/6N1/2B2Q2/8/PB3PPP/R4K1R w - - bm #9;

Result 1 min/single CPU: Rybka 2, Naum 14, cyclone 16, bright 15!

Jouni