Bad chess. Good results. Compiled from several billion regrettable decisions.
Open · Modern variation
The engine already knows how to defend. Your opponent is called rookEnjoyer94.
Masters, Mortals, and Noobs make different mistakes. We weight every reply by what that rating pool actually plays—not by what an engine demands.
On your turns, the Almanac commits to one learned move. On theirs, it keeps the messy human reply distribution. One burgundy arrow is your move; several arrows are their habits.
Expected Score asks what happens if you know our theory and they keep being human. It is not “+0.2 after perfect play.”
We replay the position against what your chosen population actually does next, then score the resulting wins, draws, and losses. On your turn we recommend; on theirs we forecast.
Every result comes from real games in the selected rating pool. A draw is zero; a win and a loss pull equally in opposite directions.
When building a finite repertoire, dangerous branches become progressively more expensive as Cowardice rises. Nothing is hard-banned; one unlikely refutation simply has to earn its risk.
At Fearless we chase the best average. Toward Fortress, replies that leave you in trouble matter more and more—even when they are not the population’s favorite.
Not engine truth: this ranks moves for the humans you selected. Stockfish has infinite patience; your opponent has three minutes and a pulse.