I made a post about a week ago sharing a chess engine I built designed to checkmate humans in as few moves as possible, rather than just playing the "best" move: https://www.reddit.com/r/ComputerChess/comments/1q34dqu/i\_built\_an\_engine\_to\_checkmate\_humans\_in\_as\_few/. In this game, the human wins if the engine doesn't checkmate them in 30 moves (30 +/- depending on the difficulty setting)

In total, 2,609 games were played to completion. Closer to 2,900 were actually started but about 300 games broke due to issues with my server (causing the game to terminate early).

I ran an A/B/C test on 464 of the games where players were randomly matched against either my engine, Stockfish 17, or Stockfish 11 with contempt=100. I ran this on games where the the engine had to get checkmate in 25 moves or more. Results:

UX Observations

• The Slider: The difficulty slider was randomized between 800-1500 Elo by default, but >50% of users never touched it. This is partially why I'm not concerned that the 44% win rate is below 50%... many players likely played the default difficulty rather than turning it up to their ELO.
• Color Balance: Users won slightly more often as White. I’ve updated the logic to require White to avoid checkmate a few more moves than before.

The page is still up if anyone else would like to try it. I appreciate all the comments in the prior thread... lots of good suggestions and questions.
http://siegechess.com