Chess Vision watches your chess.com board and quietly points out the move a strong human would play. Not the perfect computer move. The believable one. You play, it suggests, and you win the way a good player on a good day wins.
Perfect moves in a 1500 game stand out a mile. Chess Vision plays like a good human instead: a little better than the person across the board, with small, natural slips in the right places. Every single move it asks one question: is this a moment to be sharp, or am I far enough ahead to play something relaxed?
No typing moves in. It looks at your screen, recognises every piece on chess.com's classic green board, and only trusts a position once it is a real, legal move. No mix-ups while pieces are still sliding.
Stockfish calculates every position for you. The strength is always there. Chess Vision’s real job is deciding how much of it to show.
Before a move reaches your screen it passes through a model of how real players actually choose. That is what makes it believable, and it is what the rest of this page is about.
Every rating comes with its own amount of imprecision. A 1600 gives up a little ground most moves. A 2000 gives up less. Chess Vision knows that curve by heart and plays exactly on it, at whatever strength you choose.
Chess Vision looks at every decent move and gives each one a chance of being chosen, based on how much it costs. Small inaccuracies happen often, just like they do for everyone. Big ones almost never. And when there is only one move that works, a recapture, a mate in one, it plays it. Every time.
Checks are hard to resist. Captures are magnetic. A pawn about to promote is impossible to miss. Chess Vision leans toward the moves a human eye is drawn to, and away from the ones people instinctively avoid.
That last one is the classic engine idea that gets a real player’s queen trapped. Chess Vision will not play it either.
Nobody good moves a rook out and straight back. Chess Vision remembers where its pieces just came from and refuses the pointless back-and-forth that screams “bot”.
Real players do not blunder at random. They grab a poisoned pawn or attack a move too early. Now and then, and never in a critical moment, Chess Vision considers exactly those tempting moves. When it slips, it slips like a person.
Every game, it picks a comfortable lead to aim for, on average about two and a half pawns. Once you are that far ahead it eases off and plays looser, more relaxed moves, while never letting the win slip away. Some of your games will be tight. Some comfortable. None will be a demolition that looks nothing like you.
And after six perfect moves in a row it deliberately loosens up, because nobody plays six perfect moves in a row. In the background it keeps checking how strong it is really playing and corrects itself mid-game, so the strength you asked for is the strength you get.
Pick a strength and Chess Vision starts there. Then it watches how your opponent actually plays, works out how strong they really are, and settles into being just a bit better than them. Sometimes a lot better, sometimes barely. So some games are close, and once in a while you will lose one. That is exactly what makes it look real.
When you are ahead it relaxes, but never gives back more than a sliver in one move. When you are level or behind, it focuses. That is how a human who wants to win actually plays.
Pick your strength, set it up once for your board, and play.
Billed monthly. Runs on macOS 12 or later (Apple Silicon) and Windows 10/11, with chess.com open in your browser.
Get Chess Vision