Human Test

Introduction

The aim of this test was to compare in a systematic way the ability of various Chess Engines to suggest or play human-type moves (positional sacrifices being our main concern here). The test ‘jig’ we applied involved the following procedure which was carried out for each engine:

• To every engine I presented the same 25 test positions, chosen from games in which a typical ‘human’ positional sacrifice was played.
• The test consisted in having each engine ‘solve’ the position in question. If after 30 seconds the engine gives as its best move the move actually played in the game then the position was considered solved by that engine.
• What was paramount to our test was that the engine after hitting on the right move kept it as its first option up to the passing of the 30th second.
• The test would be considered negative if the engine after choosing the target move and kept as its top option for any duration of time then substituted it with some other alternative at the passing of the 30th second of analysis.
• To test the true style of play of the engines in the group access to opening books was negated.
• Allowance was made for possible solutions arrived at by transpositions as in the case shown in the following diagram.

Aronian Levon-Carlsen Magnus
Nice 2010

The move we are after here is 20… Cexd5, but the following variations are equivalent and are therefore considered both valid:









Results

Following are the results registered for the various engines:



The test confirmed our belief in Vitruvius’ ability to perform close to human standards. This is emphatically so for the Human version, but even the Conservative version’s performance is way and above that of the other engines. Only Zappa Mexico (dissident aggressor configuration) is in some ways comparable to Vitruvius.

Test Positions

FEN: 2rqkb1r/1b1n1ppp/pn1pp3/1p6/3NP1P1/2NBBP2/PPP2Q1P/1K1R3R SOLUTION: 1… Txc3	FEN: r4rk1/2q1bppp/p1b2n2/1p2p3/1P2P3/P1NBB3/2P1QRPP/5RK1 SOLUTION: 1. Txf6
FEN: r1bqr1k1/pp3pbp/2n2np1/2Pp4/3N4/2N1BP2/PP1QB1PP/R4RK1 SOLUTION: 1… Txe3	FEN: b2r3r/k4p1p/p2q1np1/NppP4/3p1Q2/P4PPB/1PP4P/1K1RR3 SOLUTION: 1. Txd4
FEN: 1q6/2Rb3k/4rpp1/1p2p2p/pP1rP2P/P3QPP1/2R4K/5B2 SOLUTION: 1. Dxd4	FEN: 1r3rk1/p4ppp/b1p1p3/2qn4/2P5/1Q4P1/P2BPPBP/R2R2K1 SOLUTION: 1. cxd5
FEN: r4bk1/p1B2pp1/1pp3p1/2P1R1P1/2nP2B1/P6P/5P2/6K1 SOLUTION: 1. d5	FEN: r2q1nk1/3b1rb1/N2p4/3Pp1pp/p3Pp2/5P2/1P2BBPP/2RQ1R1K SOLUTION: 1. Tc6
FEN: 2r1kb1r/1b1n1ppp/pq1ppn2/1p4B1/4PP2/1NNB3Q/PPP3PP/2KRR3 SOLUTION: 1… Txc3	FEN: 2r2rk1/2pR1p1b/1pq1pQp1/p3P2p/P1PR3P/1P3P2/6PK/8 SOLUTION: 1. Rg3
FEN: r2q1rk1/p2pbppp/1pn5/2p2N2/1n6/6P1/PP1QPPBP/R1B1K2R b KQ SOLUTION: 1… Cd4	FEN: 4r1k1/pp2pp1p/3p2p1/2qb1PPn/3N4/2P1R3/P1P2Q1R/2K5 SOLUTION: 1. Txh5
FEN: r1r3k1/1b2qnpp/p2b1p2/Pp1p4/3P1P1P/1Q2P1P1/1B1NBK2/R2R4 SOLUTION: 1… Tc4	FEN: 8/p3q1kp/1p2Pnp1/3pQ3/2pP4/1nP3N1/1B4PP/6K1 SOLUTION: 1. Aa3
FEN: r1b4r/1p2kppp/3Npn2/2q1p3/8/2N4P/PPP3P1/R2Q1R1K SOLUTION: 1. Txf6	FEN: 3r2k1/1r1qpp1p/2R3p1/1p1P4/nP1Q3P/6P1/5PB1/4R1K1 SOLUTION: 1. h5
FEN: r3r1k1/ppqnbppp/2p2n2/5N2/2P5/2B1PN2/PPQ3PP/2KR3R SOLUTION: 1. Cxg7	FEN: r2q1rk1/1b3ppp/p4n2/3P4/1bpQ4/5N1P/P1B2PP1/R1B1R1K1 SOLUTION: 1. Tb1
FEN: 1rq1kbn1/3b1ppr/p2p3p/2nPp2N/4P2B/N4Q2/2R1BPPP/2R3K1 SOLUTION: 1. Txc5	FEN: rq3rk1/p2p2pp/1pb1p3/4b2Q/2P1PB2/P2B2P1/1P5P/2KR1R2 SOLUTION: 1… Txf4
FEN: r4b1k/3b2r1/pq1p1n1n/3Pp1pp/1P2Pp2/N1R2P1P/1Q1BBNP1/2R3K1 SOLUTION: 1. Tc6	FEN: 3rq1k1/4rppp/2n3b1/pp2P3/2pP1QB1/P1P1R3/1B4PP/4R1K1 SOLUTION: 1… Te6
FEN: r3qrk1/1pp4p/5p1P/p1p1p3/P3Pn2/1P1P4/1BP4Q/R4RK1 SOLUTION: 1. Txf4	FEN: 4r1k1/1p1qppbp/p2p1np1/P2P1b2/Rnr5/1NN3P1/1PPQ1PBP/2B1R1K1 SOLUTION: 1… Txc3
FEN: r2q1r1k/1b2n1bp/3p1np1/p2Pp3/NpP3P1/1P6/P2BBNKP/2RQ1R2 SOLUTION: 1… Cexd5