Artemis Chess Engine

Artemis CE is a cutting-edge chess engine currently under development, with a strong emphasis on performance and innovative optimization techniques. The engine's move generation is now fully complete and highly efficient (see benchmarks below). Like most modern chess engines, Artemis CE uses a bitboard representation: each piece type is stored as a 64-bit integer, with one bit corresponding to each square on the board. This compact format enables extremely fast computation of attacks and legal moves through bitwise operations (AND, OR, shifts, etc.). As development continues, Artemis CE will focus on pioneering new algorithms in move generation, search, and evaluation to push the boundaries of chess engine performance.

Perft Benchmarks

Perft (performance test, move path enumeration) is a standard benchmarking method for chess engines. It measures an engine's move generation speed (and correctness) by counting all legal move sequences to a specified depth in a given position. In these single-threaded benchmarks, Artemis CE outperformed Stockfish 17 in 6 out of 7 positions, with particularly strong gains in open positions where sliding pieces have greater mobility. In such cases, the bitboards representations for sliding piece move sets are denser, allowing Artemis CE's move set hashing optimizations to shine.

Position 1

time to generate 13.8B moves

Artemis 67 s

Stockfish 115 s

Position 2

time to generate 4.7B moves

Artemis 21 s

Stockfish 36 s

Position 3

time to generate 4.9B moves

Artemis 22 s

Stockfish 34 s

Position 4

time to generate 26.3B moves

Artemis 89 s

Stockfish 111 s

Position 5

time to generate 1.7B moves

Artemis 7.7 s

Stockfish 8.3 s

Position 6

time to generate 51.1B moves

Artemis 205 s

Stockfish 211 s

Position 7

time to generate 84.9B moves

Artemis 429 s

Stockfish 397 s

Benchmark Methodology: All tests were performed on an AMD Ryzen 9 3950X 16-Core Processor using release builds of Stockfish 17 and Artemis CE in single-threaded mode. For each position, perft (move path enumeration) times were measured at the same depth for both engines to ensure a direct and fair performance comparison. Note that the selected depth varies across positions and was chosen individually to demonstrate meaningful differences in move generation efficiency.