This is a chess testset in EPD format composed of puzzles considered very hard for chess engines. The aim is not testing the engines, but to collect and preserve positions that humans but not engines where able to solve (in reasonable time).

So this test set will:

  • Contain only positions from human games or human studies (with reference to the source)

  • Be sound, so that between best move(s) and the others there is an important difference

  • Be really hard to solve

For each position there will be a leaderboard with the name of who found the solution / hardware and software used / PV line with depth and time.

The list will grow with time, but the size of the list is not a point, the difficult of the puzzles and the reference to the sources are more important, so the list will be extremely selective.

Positions are collected from different sources, the origin of the position and eveltually the original testset will always be kept and granted the due credit, indeed anonymous positions are not allowed.

Everybody is welcome for help and support both as providing new puzzles or helping testing the suite.

View the Project on GitHub