Game database
Every finished live game of Coherent, with every move, the engines’ evaluation and the clocks. Find a game, then go through it at your own pace.
These games are no register runs and no measurement of playing strength. Live →
Every finished live game of Coherent, with every move, the engines’ evaluation and the clocks. Find a game, then go through it at your own pace.
These games are no register runs and no measurement of playing strength. Live →
Several words narrow the list; small typos are forgiven.
Loading the games …
Choose a game from the list.
← Black betterWhite better →
Evaluations from White’s point of view, in pawns; # means mate in so many moves.
Searched in the game
Download PGN · Link to this position
Arrow keys step through the moves, Home and End jump to start and end, F flips the board.
Two engines are drawn per game from a pool of free top engines (at present Stockfish 19, Obsidian 16 and Reckless 0.9), with endgame tablebases up to six pieces. Each of the two searches every position in which a playing engine chose its move itself (not the opening moves set by the book, not moves without an alternative) for twice the time that player took for the move, on a processor core of its own.
Only when both engines agree. The measure is the expected score a move gives away under Stockfish 19’s win/draw/loss model: near equality a game between top engines stays drawn, it tips only from about a pawn’s advantage. 5, 10 and 20 percentage points lost (and at least 25, 40 and 60 centipawns) make an inaccuracy, a mistake and a blunder; nothing is marked in a position already lost. The curve shows White’s expected score through the game.
Not every game is analysed: decisive games first, newest first; draws when no decisive game waits. A game in which the time of a move was not recorded is left out. A deep analysis, not a measurement of playing strength.