Proving the correctness of the algorithm

[Fguy64]

I am looking for some baseline evaluation numbers that I can use to validate my algorithm

for example, a program that implements, at any ply, only a negamax search, with alpha-beta pruning, no quiescence (that is gonna have to wait), no hashtable, and an evaluation based only on material considerations, with standard set of piece values, that program is either correct or incorrect, and any two correct implementations should give identical results.

Agreed?

So with that in mind, any suggestions for pursuing this angle of comparison to a standard set of evaluations will be much appreciated.

Thanks,
Fred.

[Aleks Peshkov]

Alpha-beta is dependent to move-ordering, so slightly different programs will give different results.

Most will agree that 1-4-4-6-12 piece values works better then standard human values.

Piece-square tables (PST) positional evaluation is trivial but gives decent playing style.

A competitive chess program should know and avoid 3-fold (2-fold) repetitions draw rule when it ahead in material.

[Fguy64]

hmm ok yes thanks, I can see that order of possible moves at a given ply can cause alphabeta pruning to kick in at different times and provide different evaluations, so I will qualify my original hypothesis by adding "and the same move order" will result in identical results.

Anyways, thanks for the remarks, I'm still hoping that standard baseline evals should still be available (I hope) to validate my algorithm.

[Sven]

Alpha-beta is dependent to move-ordering, so slightly different programs will give different results.

Most will agree that 1-4-4-6-12 piece values works better then standard human values.

Piece-square tables (PST) positional evaluation is trivial but gives decent playing style.

A competitive chess program should know and avoid 3-fold (2-fold) repetitions draw rule when it ahead in material.

I think Fred wants to do something where neither move ordering nor positional evaluation is relevant. I think he only wants to test the correctness of his alpha-beta negamax function. So move ordering will not affect the result which is simply a number (the material evaluation of the PV his search function finds) but only the efficiency, assuming a fixed-depth search of course.

I think one small issue of this approach in practice may be to find a "reference engine" where quiescence search, hash table, null move, positional evaluation, extensions, reductions, and other features are either not implemented or can be switched off easily. Such an engine can be written, of course, but then we need to prove somehow the correctness of *this* engine, which may or may not be helpful for Fred.

Whether the "reference engine" implements repetition detection or not might be subject to definition. The same applies for other kinds of draw detection.

Sven

[Fguy64]

Alpha-beta is dependent to move-ordering, so slightly different programs will give different results.

Most will agree that 1-4-4-6-12 piece values works better then standard human values.

Piece-square tables (PST) positional evaluation is trivial but gives decent playing style.

A competitive chess program should know and avoid 3-fold (2-fold) repetitions draw rule when it ahead in material.

I think Fred wants to do something where neither move ordering nor positional evaluation is relevant. I think he only wants to test the correctness of his alpha-beta negamax function. So move ordering will not affect the result which is simply a number (the material evaluation of the PV his search function finds) but only the efficiency, assuming a fixed-depth search of course.

I think one small issue of this approach in practice may be to find a "reference engine" where quiescence search, hash table, null move, positional evaluation, extensions, reductions, and other features are either not implemented or can be switched off easily. Such an engine can be written, of course, but then we need to prove somehow the correctness of *this* engine, which may or may not be helpful for Fred.

Whether the "reference engine" implements repetition detection or not might be subject to definition. The same applies for other kinds of draw detection.

Sven

I'm not sure I follow your logic. Yes you are correct I would like to test the correctness of my negamax algorithm, and if I could test that both with and without alpha-beta that would be a bonus. But I would say that move order is relevant with alpha-beta which seems to me to be at odds with the statement I have bolded. It is my understanding that move order is irrelevant only when you have negamax-without alpha-beta. see my point?

Is this just semantics here? I think I agree that that eval of the final move chosen should not be affected by alpha-beta, but I'd still like to verify the evals of the candidates not chosen, and that is where move ordering would have a bearing. agreed? I'm just trying to nail down my semantics here.

[Sven]

I think I agree that that eval of the final move chosen should not be affected by alpha-beta, but I'd still like to verify the evals of the candidates not chosen, and that is where move ordering would have a bearing. agreed?

O.k., you didn't tell that before. Your initial posting gave the impression that you only wanted to compare the eval of the PV with that of a reference engine. In this case the resulting eval of a deterministic fixed-depth search must always be the same, with or without alpha-beta, with move ordering strategy A, B, or C. You always get the minimax value of the position, sooner or later. Even with qsearch. Even with positional eval.

Now that you want to include other root moves, too, it will indeed be necessary to define exactly how the "reference engine" implements move ordering, and you'll have to follow that in your own search. This might include the order of move generation, too, so it gets even more difficult to achieve a perfect "standard".

In my opinion you still have the option to return to the "eval of PV only" idea which might save you a lot of trouble. Also I see no real benefit in looking at evaluations of moves other than the best move, since in reality many of these values are only upper bounds. It has zero impact on correctness whether you find the *best* refutation of a non-optimal move, or just a *sufficient* one that triggers a cutoff.

Sven

[jwes]

I'm not sure if this is what you are looking for, but this is a very easy test set (they are the easiest positions from WAC). Any engine with a working search and material evaluation should get them all right. Crafty finds the "hardest" in 6 ply.

5rk1/1ppb3p/p1pb4/6q1/3P1p1r/2P1R2P/PP1BQ1P1/5RKN w - - bm Rg3; id "WAC.003";
r1bq2rk/pp3pbp/2p1p1pQ/7P/3P4/2PB1N2/PP3PPR/2KR4 w - - bm Qxh7+; id "WAC.004";
5k2/6pp/p1qN4/1p1p4/3P4/2PKP2Q/PP3r2/3R4 b - - bm Qc4+; id "WAC.005";
7k/p7/1R5K/6r1/6p1/6P1/8/8 w - - bm Rb7; id "WAC.006";
rnbqkb1r/pppp1ppp/8/4P3/6n1/7P/PPPNPPP1/R1BQKBNR b KQkq - bm Ne3; id "WAC.007";
r4q1k/p2bR1rp/2p2Q1N/5p2/5p2/2P5/PP3PPP/R5K1 w - - bm Rf7; id "WAC.008";
3q1rk1/p4pp1/2pb3p/3p4/6Pr/1PNQ4/P1PB1PP1/4RRK1 b - - bm Bh2+; id "WAC.009";
2br2k1/2q3rn/p2NppQ1/2p1P3/Pp5R/4P3/1P3PPP/3R2K1 w - - bm Rxh7; id "WAC.010";
r1b1kb1r/3q1ppp/pBp1pn2/8/Np3P2/5B2/PPP3PP/R2Q1RK1 w kq - bm Bxc6; id "WAC.011";
4k1r1/2p3r1/1pR1p3/3pP2p/3P2qP/P4N2/1PQ4P/5R1K b - - bm Qxf3+; id "WAC.012";
r2rb1k1/pp1q1p1p/2n1p1p1/2bp4/5P2/PP1BPR1Q/1BPN2PP/R5K1 w - - bm Qxh7+; id "WAC.014";
1R6/1brk2p1/4p2p/p1P1Pp2/P7/6P1/1P4P1/2R3K1 w - - bm Rxb7; id "WAC.015";
r2qkb1r/1ppb1ppp/p7/4p3/P1Q1P3/2P5/5PPP/R1B2KNR b kq - bm Bb5; id "WAC.020";
6k1/1b1nqpbp/pp4p1/5P2/1PN5/4Q3/P5PP/1B2B1K1 b - - bm Bd4; id "WAC.024";
3R1rk1/8/5Qpp/2p5/2P1p1q1/P3P3/1P2PK2/8 b - - bm Qh4+; id "WAC.025";
7k/pp4np/2p3p1/3pN1q1/3P4/Q7/1r3rPP/2R2RK1 w - - bm Qf8+; id "WAC.027";
1r1r2k1/4pp1p/2p1b1p1/p3R3/RqBP4/4P3/1PQ2PPP/6K1 b - - bm Qe1+; id "WAC.028";
r2q2k1/pp1rbppp/4pn2/2P5/1P3B2/6P1/P3QPBP/1R3RK1 w - - bm c6; id "WAC.029";
7k/1b1r2p1/p6p/1p2qN2/3bP3/3Q4/P5PP/1B1R3K b - - bm Bg1; id "WAC.034";
r3r2k/2R3pp/pp1q1p2/8/3P3R/7P/PP3PP1/3Q2K1 w - - bm Rxh7+; id "WAC.035";
3r4/2p1rk2/1pQq1pp1/7p/1P1P4/P4P2/6PP/R1R3K1 b - - bm Re1+; id "WAC.036";
r1br2k1/pp2bppp/2nppn2/8/2P1PB2/2N2P2/PqN1B1PP/R2Q1R1K w - - bm Na4; id "WAC.039";
3r1r1k/1p4pp/p4p2/8/1PQR4/6Pq/P3PP2/2R3K1 b - - bm Rc8; id "WAC.040";
3rb1k1/pq3pbp/4n1p1/3p4/2N5/2P2QB1/PP3PPP/1B1R2K1 b - - bm dxc4; id "WAC.044";
7k/2p1b1pp/8/1p2P3/1P3r2/2P3Q1/1P5P/R4qBK b - - bm Qxa1; id "WAC.045";
k4r2/1R4pb/1pQp1n1p/3P4/5p1P/3P2P1/r1q1R2K/8 w - - bm Rxb6+; id "WAC.050";
r1bq1r2/pp4k1/4p2p/3pPp1Q/3N1R1P/2PB4/6P1/6K1 w - - bm Rg4+; id "WAC.051";
r1k5/1p3q2/1Qpb4/3N1p2/5Pp1/3P2Pp/PPPK3P/4R3 w - - bm Re7; id "WAC.052";
6k1/6p1/p7/3Pn3/5p2/4rBqP/P4RP1/5QK1 b - - bm Re1; id "WAC.053";
r3kr2/1pp4p/1p1p4/7q/4P1n1/2PP2Q1/PP4P1/R1BB2K1 b q - bm Qh1+; id "WAC.054";
r1bqk2r/pppp1ppp/5n2/2b1n3/4P3/1BP3Q1/PP3PPP/RNB1K1NR b KQkq - bm Bxf2+; id "WAC.056";
r3q1kr/ppp5/3p2pQ/8/3PP1b1/5R2/PPP3P1/5RK1 w - - bm Rf8+; id "WAC.057";
r1b2rk1/2p1qnbp/p1pp2p1/5p2/2PQP3/1PN2N1P/PB3PP1/3R1RK1 w - - bm Nd5; id "WAC.059";
rn1qr1k1/1p2np2/2p3p1/8/1pPb4/7Q/PB1P1PP1/2KR1B1R w - - bm Qh8+; id "WAC.060";
3qrbk1/ppp1r2n/3pP2p/3P4/2P4P/1P3Q2/PB6/R4R1K w - - bm Qf7+; id "WAC.061";
r1brnbk1/ppq2pp1/4p2p/4N3/3P4/P1PB1Q2/3B1PPP/R3R1K1 w - - bm Nxf7; id "WAC.063";
8/6pp/3q1p2/3n1k2/1P6/3NQ2P/5PP1/6K1 w - - bm g4+; id "WAC.064";
1r1r1qk1/p2n1p1p/bp1Pn1pQ/2pNp3/2P2P1N/1P5B/P6P/3R1RK1 w - - bm Ne7+; id "WAC.065";
3r2k1/p2q4/1p4p1/3rRp1p/5P1P/6PK/P3R3/3Q4 w - - bm Rxd5; id "WAC.067";
6k1/5ppp/1q6/2b5/8/2R1pPP1/1P2Q2P/7K w - - bm Qxe3; id "WAC.068";
2kr3r/pppq1ppp/3p1n2/bQ2p3/1n1PP3/1PN1BN1P/1PP2PP1/2KR3R b - - bm Na2+; id "WAC.070";
r3r1k1/pp1n1ppp/2p5/4Pb2/2B2P2/B1P5/P5PP/R2R2K1 w - - bm e6; id "WAC.072";
r1q3rk/1ppbb1p1/4Np1p/p3pP2/P3P3/2N4R/1PP1Q1PP/3R2K1 w - - bm Qd2; id "WAC.073";
5r1k/pp4pp/2p5/2b1P3/4Pq2/1PB1p3/P3Q1PP/3N2K1 b - - bm Qf1+; id "WAC.074";
r3r1k1/pppq1ppp/8/8/1Q4n1/7P/PPP2PP1/RNB1R1K1 b - - bm Qd6; id "WAC.075";
3r2k1/ppp2ppp/6q1/b4n2/3nQB2/2p5/P4PPP/RN3RK1 b - - bm Ng3; id "WAC.077";
r2q3r/ppp2k2/4nbp1/5Q1p/2P1NB2/8/PP3P1P/3RR1K1 w - - bm Ng5+; id "WAC.078";
r3k2r/pbp2pp1/3b1n2/1p6/3P3p/1B2N1Pq/PP1PQP1P/R1B2RK1 b kq - bm Qxh2+; id "WAC.079";
r2q1r1k/2p1b1pp/p1n5/1p1Q1bN1/4n3/1BP1B3/PP3PPP/R4RK1 w - - bm Qg8+; id "WAC.084";
kr2R3/p4r2/2pq4/2N2p1p/3P2p1/Q5P1/5P1P/5BK1 w - - bm Na6; id "WAC.085";
r6k/p1Q4p/2p1b1rq/4p3/B3P3/4P3/PPP3P1/4RRK1 b - - bm Rxg2+; id "WAC.088";
8/k7/p7/3Qp2P/n1P5/3KP3/1q6/8 b - - bm e4+; id "WAC.094";
2r5/1r6/4pNpk/3pP1qp/8/2P1QP2/5PK1/R7 w - - bm Ng4+; id "WAC.095";
6k1/5p2/p5np/4B3/3P4/1PP1q3/P3r1QP/6RK w - - bm Qa8+; id "WAC.097";
r1bq1r1k/1pp1Np1p/p2p2pQ/4R3/n7/8/PPPP1PPP/R1B3K1 w - - bm Rh5; id "WAC.099";
2Q2n2/2R4p/1p1qpp1k/8/3P3P/3B2P1/5PK1/r7 w - - bm Qxf8+; id "WAC.102";
6k1/2pb1r1p/3p1PpQ/p1nPp3/1q2P3/2N2P2/PrB5/2K3RR w - - bm Qxg6+; id "WAC.103";
b4r1k/pq2rp2/1p1bpn1p/3PN2n/2P2P2/P2B3K/1B2Q2N/3R2R1 w - - bm Qxh5; id "WAC.104";
4rrk1/pppb4/7p/3P2pq/3Qn3/P5P1/1PP4P/R3RNNK b - - bm Nf2+; id "WAC.106";
5n2/pRrk2p1/P4p1p/4p3/3N4/5P2/6PP/6K1 w - - bm Nb5; id "WAC.107";
2kr4/bp3p2/p2p2b1/P7/2q5/1N4B1/1PPQ2P1/2KR4 b - - bm Be3; id "WAC.110";
6k1/p5p1/5p2/2P2Q2/3pN2p/3PbK1P/7P/6q1 b - - bm Qf1+; id "WAC.111";
r4kr1/ppp5/4bq1b/7B/2PR1Q1p/2N3P1/PP3P1P/2K1R3 w - - bm Rxe6; id "WAC.112";
3r1rk1/q4ppp/p1Rnp3/8/1p6/1N3P2/PP3QPP/3R2K1 b - - bm Ne4; id "WAC.117";
6k1/5p1p/2bP2pb/4p3/2P5/1p1pNPPP/1P1Q1BK1/1q6 b - - bm Bxf3+; id "WAC.121";
1k6/ppp4p/1n2pq2/1N2Rb2/2P2Q2/8/P4KPP/3r1B2 b - - bm Rxf1+; id "WAC.122";
r1bqr1k1/pp3ppp/1bp5/3n4/3B4/2N2P1P/PPP1B1P1/R2Q1RK1 b - - bm Bxd4+; id "WAC.125";
r5r1/pQ5p/1qp2R2/2k1p3/4P3/2PP4/P1P3PP/6K1 w - - bm Rxc6+; id "WAC.126";
3r1r1k/1b2b1p1/1p5p/2p1Pp2/q1B2P2/4P2P/1BR1Q2K/6R1 b - - bm Bf3; id "WAC.129";
3r2k1/p6p/2Q3p1/4q3/2P1p3/P3Pb2/1P3P1P/2K2BR1 b - - bm Rd1+; id "WAC.134";
6kr/1q2r1p1/1p2N1Q1/5p2/1P1p4/6R1/7P/2R3K1 w - - bm Rc8+; id "WAC.136";
rnb3kr/ppp2ppp/1b6/3q4/3pN3/Q4N2/PPP2KPP/R1B1R3 w - - bm Nf6+; id "WAC.139";
5b2/pp2r1pk/2pp1pRp/4rP1N/2P1P3/1P4QP/P3q1P1/5R1K w - - bm Rxh6+; id "WAC.143";
r1b2rk1/2p2ppp/p7/1p6/3P3q/1BP3bP/PP3QP1/RNB1R1K1 w - - bm Qxf7+; id "WAC.154";
r1b1qN1k/1pp3p1/p2p3n/4p1B1/8/1BP4Q/PP3KPP/8 w - - bm Qxh6+; id "WAC.156";
5rk1/n1p1R1bp/p2p4/1qpP1QB1/7P/2P3P1/PP3P2/6K1 w - - bm Rxg7+; id "WAC.158";
r1b2r2/5P1p/ppn3pk/2p1p1Nq/1bP1PQ2/3P4/PB4BP/1R3RK1 w - - bm Ne6+; id "WAC.159";
qn1kr2r/1pRbb3/pP5p/P2pP1pP/3N1pQ1/3B4/3B1PP1/R5K1 w - - bm Qxd7+; id "WAC.160";
3r3k/3r1P1p/pp1Nn3/2pp4/7Q/6R1/Pq4PP/5RK1 w - - bm Qxd8+; id "WAC.161";
8/6pp/4p3/1p1n4/1NbkN1P1/P4P1P/1PR3K1/r7 w - - bm Rxc4+; id "WAC.164";
r3r1k1/5pp1/p1p4p/2Pp4/8/q1NQP1BP/5PP1/4K2R b K - bm d4; id "WAC.166";
r3k2r/pb1q1p2/8/2p1pP2/4p1p1/B1P1Q1P1/P1P3K1/R4R2 b kq - bm Qd2+; id "WAC.168";
5rk1/1pp3bp/3p2p1/2PPp3/1P2P3/2Q1B3/4q1PP/R5K1 b - - bm Bh6; id "WAC.169";
5r1k/6Rp/1p2p3/p2pBp2/1qnP4/4P3/Q4PPP/6K1 w - - bm Qxc4; id "WAC.170";
5r1k/p5pp/8/1P1pq3/P1p2nR1/Q7/5BPP/6K1 b - - bm Qe1+; id "WAC.172";
2r1b3/1pp1qrk1/p1n1P1p1/7R/2B1p3/4Q1P1/PP3PP1/3R2K1 w - - bm Qh6+; id "WAC.173";
r5k1/pppb3p/2np1n2/8/3PqNpP/3Q2P1/PPP5/R4RK1 w - - bm Nh5; id "WAC.175";
r1b3r1/4qk2/1nn1p1p1/3pPp1P/p4P2/1p3BQN/PKPBN3/3R3R b - - bm Qa3+; id "WAC.177";
r1b2r1k/pp4pp/3p4/3B4/8/1QN3Pn/PP3q1P/R3R2K b - - bm Qg1+; id "WAC.179";
r3k2r/2p2p2/p2p1n2/1p2p3/4P2p/1PPPPp1q/1P5P/R1N2QRK b kq - bm Ng4; id "WAC.181";
4kn2/r4p1r/p3bQ2/q1nNP1Np/1p5P/8/PPP3P1/2KR3R w - - bm Qe7+; id "WAC.184";
r5r1/p1q2p1k/1p1R2pB/3pP3/6bQ/2p5/P1P1NPPP/6K1 w - - bm Bf8+; id "WAC.186";
3RNbk1/pp3p2/4rQpp/8/1qr5/7P/P4P2/3R2K1 w - - bm Qg7+; id "WAC.188";
3r1k2/1ppPR1n1/p2p1rP1/3P3p/4Rp1N/5K2/P1P2P2/8 w - - bm Re8+; id "WAC.189";
2r1Rn1k/1p1q2pp/p7/5p2/3P4/1B4P1/P1P1QP1P/6K1 w - - bm Qc4; id "WAC.191";
r3k3/ppp2Npp/4Bn2/2b5/1n1pp3/N4P2/PPP3qP/R2QKR2 b Qq - bm Nd3+; id "WAC.192";
7k/1p4p1/7p/3P1n2/4Q3/2P2P2/PP3qRP/7K b - - bm Qf1+; id "WAC.197";
2b2r1k/4q2p/3p2pQ/2pBp3/8/6P1/1PP2P1P/R5K1 w - - bm Ra7; id "WAC.201";
r4rk1/5ppp/p3q1n1/2p2NQ1/4n3/P3P3/1B3PPP/1R3RK1 w - - bm Qh6; id "WAC.203";
4kb1r/2q2p2/r2p4/pppBn1B1/P6P/6Q1/1PP5/2KRR3 w k - bm Rxe5+; id "WAC.209";
r1bqrk2/pp1n1n1p/3p1p2/P1pP1P1Q/2PpP1NP/6R1/2PB4/4RBK1 w - - bm Qxf7+; id "WAC.211";
rn1qr2Q/pbppk1p1/1p2pb2/4N3/3P4/2N5/PPP3PP/R4RK1 w - - bm Qxg7+; id "WAC.212";
3r2k1/pb1q1pp1/1p2pb1p/8/3N4/P2QB3/1P3PPP/1Br1R1K1 w - - bm Qh7+; id "WAC.215";
r3kb1r/1pp3p1/p3bp1p/5q2/3QN3/1P6/PBP3P1/3RR1K1 w kq - bm Qd7+; id "WAC.217";
6k1/pp5p/2p3q1/6BP/2nPr1Q1/8/PP3R1K/8 w - - bm Bh6; id "WAC.218";
7k/p4q1p/1pb5/2p5/4B2Q/2P1B3/P6P/7K b - - bm Qf1+; id "WAC.219";
3rr1k1/ppp2ppp/8/5Q2/4n3/1B5R/PPP1qPP1/5RK1 b - - bm Qxf1+; id "WAC.220";
r3k3/P5bp/2N1bp2/4p3/2p5/6NP/1PP2PP1/3R2K1 w q - bm Rd8+; id "WAC.221";
4R3/4q1kp/6p1/1Q3b2/1P1b1P2/6KP/8/8 b - - bm Qh4+; id "WAC.225";
5rk1/p1p2r1p/2pp2p1/4p3/PPPnP3/3Pq1P1/1Q1R1R1P/4NK2 b - - bm Nb3; id "WAC.233";
2kr1r2/p6p/5Pp1/2p5/1qp2Q1P/7R/PP6/1KR5 w - - bm Rb3; id "WAC.234";
1r3r1k/3p4/1p1Nn1R1/4Pp1q/pP3P1p/P7/5Q1P/6RK w - - bm Qe2; id "WAC.243";
r6r/pp3ppp/3k1b2/2pb4/B4Pq1/2P1Q3/P5PP/1RBR2K1 w - - bm Qxc5+; id "WAC.244";
6R1/4qp1p/ppr1n1pk/8/1P2P1QP/6N1/P4PP1/6K1 w - - bm Qh5+; id "WAC.246";
k5r1/p4b2/2P5/5p2/3P1P2/4QBrq/P5P1/4R1K1 w - - bm Qe8+; id "WAC.253";
r6k/pp3p1p/2p1bp1q/b3p3/4Pnr1/2PP2NP/PP1Q1PPN/R2B2RK b - - bm Nxh3; id "WAC.254";
rnbqr2k/pppp1Qpp/8/b2NN3/2B1n3/8/PPPP1PPP/R1B1K2R w KQ - bm Qg8+; id "WAC.263";
2r1kb1r/pp3ppp/2n1b3/1q1N2B1/1P2Q3/8/P4PPP/3RK1NR w Kk - bm Nc7+; id "WAC.267";
2r3kr/ppp2n1p/7B/5q1N/1bp5/2Pp4/PP2RPPP/R2Q2K1 w - - bm Re8+; id "WAC.268";
2k4B/bpp1qp2/p1b5/7p/1PN1n1p1/2Pr4/P5PP/R3QR1K b - - bm Ng3+; id "WAC.273";
r5k1/pp1RR1pp/1b6/6r1/2p5/B6P/P4qPK/3Q4 w - - bm Qd5+; id "WAC.276";
r2qkb1r/pppb2pp/2np1n2/5pN1/2BQP3/2N5/PPP2PPP/R1B1K2R w KQkq - bm Bf7+; id "WAC.278";
2rr3k/1b2bppP/p2p1n2/R7/3P4/1qB2P2/1P4Q1/1K5R w - - bm Qxg7+; id "WAC.285";
2k2r2/2p5/1pq5/p1p1n3/P1P2n1B/1R4Pp/2QR4/6K1 b - - bm Ne2+; id "WAC.290";
4r3/p4r1p/R1p2pp1/1p1bk3/4pNPP/2P1K3/2P2P2/3R4 w - - bm Rxd5+; id "WAC.295";
3Q4/p3b1k1/2p2rPp/2q5/4B3/P2P4/7P/6RK w - - bm Qh8+; id "WAC.298";

[Fguy64]

I'm not sure if this is what you are looking for, but this is a very easy test set (they are the easiest positions from WAC). Any engine with a working search and material evaluation should get them all right. Crafty finds the "hardest" in 6 ply.

....

Thanks, but as soon as you mentioned the words "find" and "hardest", then no, it doesn't seem to be what I am looking for. You see, I have no positional eval to speak of, all my eval does is add up material, and solve checkmates, nothing else. Correct me if I'm wrong, but I think your post has more to it that that?

But I will file this for use as I develop my evaluation routine, but that won't happen until I verify my negamax with alphabeta etc.

Anyways, as for my goals in this thread, I suppose there are a few things I can do on my own, for example, If I set the ply to 4, then it should solve any mate in 2, set the ply to 6 and it should solve mate in three, etc. That should go a way towards "proving" my algorithm

regards.

[Aaron Becker]

You don't need positional knowledge to solve any of those problems. As long as you're counting material correctly and searching deeply enough (apparently 6 plies should suffice for those problems), you will find the solutions.

You really don't need a sophisticated evaluation function to make a reasonably strong engine. I'm working on an engine that currently uses only material and piece square tables in its evaluation, and it scores 291/300 on WAC. Of course this is with a hash table and a more sophisticated search than plain alpha-beta, but these problems should all be solvable if your implementation is correct.

[Fguy64]

Duly noted Aaron, thanks for the clarification