Chess Moment · Fair play

Fair play, stated plainly.

We do not claim that engine assistance is impossible. We claim what we actually do — and we would rather you read it than trust a badge.

What we claim

While a live rated game is in progress, analysis, hints, the engine panel, import and the Coach are switched off for that game in every client, and the server refuses analysis requests for any position of a game you are still playing — so a second device or browser tab cannot serve as the engine. Rated play refuses hints and takebacks on the server. While any online game is live, the iOS and Android apps also keep the board out of screenshots and screen recordings; a browser cannot be stopped from capturing its own tab. The block observes nothing and reports nothing.

Nothing more is claimed. An external engine cannot be blocked. Detection plus human review is the real control, and this page says what that means.

What is measured

Only rated games between two people, and only from data the server already holds: how often your moves match the engine’s first and top-three choices and the average centipawn loss over your scored moves (opening-book moves, single legal moves, obvious only-moves and already-decided positions are excluded); how your thinking time is distributed across forced, capture, quiet and critical moves, taken from the server clock; and how often your connection dropped while your own clock was running.

At review time a person also reads your rating history, reports about you, and whether the same device runs other accounts. There is no device-side monitoring of any kind: no focus or window tracking, no keystrokes, no screen. No thresholds are published.

What is never judged

AI games, practice games against a house player, friend games, local games and aborted games produce no fair-play data at all, and a game with fewer than twenty scored moves is not scored. Casual play is never judged.

Only a person restricts an account

Automation opens a case; a reviewer decides. Three different people reporting the same account within thirty days pin its case to the top of the queue — a single report never opens one.

The ladder is a warning, then rated play paused for thirty days (ninety the second time), then a closed account; a third rated restriction closes the account. Every sanction comes with a statement of reasons in the app: the effect and its duration, the games and facts relied on, whether automated means were used and how, the clause of the Terms it rests on, and how to appeal.

Refunds

When a case is upheld, opponents from the sanctioned player’s last forty rated games who lost rating points get them back — capped at 150 and never more than they lost. Everyone who reported the account is told whether action was taken or no violation was found.

Appeals

One appeal per sanction, free, open for six months. A person decides within fourteen days, after at least forty-eight hours, with the automated inputs re-run. An overturn lifts the restriction and restores your rating; opponents’ refunds are not reversed. Until a second reviewer exists, the reply says that the same reviewer decided.

Your data

Signal rows are deleted ninety days after the game unless a reviewer has opened a case, in which case they live with the case, twenty-four months after it closes. A sanction record and its appeal stay for the life of the account; deleting the account deletes them. Pro is never revoked by a sanction — a closed account’s notice explains how billing stops in the store it was bought from. The privacy policy names the same signals and bounds.

Related: the privacy policy's retention section names the same signals and bounds, and Terms of Use §3 is the contractual ground of every sanction.