Chess
Fair play

Trust ratings

Every finished rated game is analyzed move-by-move. We compare each player's stats to typical play at their Elo and surface a trust score you can check before you sit down — or after.

How it works

When a game ends, moves are sent through Stockfish analysis. We classify each move (best, good, inaccuracy, mistake, blunder) and compute Chess.com-style CAPS2 accuracy — win-probability based, usually between 50 and 95.

Trust is not a cheat detector on its own. It measures how closely someone's recent games look like normal play at their rating: accuracy, blunder rate, mistake gaps, accuracy streaks, and instant-win patterns.

Ratings below 1600 Elo get slightly stricter scrutiny because low-rated accounts with engine-like stats are more common.

Score bands
Shown on profiles and next to names in live games.
  • Trusted85–100%
  • Normal70–84%
  • Caution50–69%
  • Suspicious30–49%
  • Highly Suspicious0–29%

Overwatch convictions cut the displayed trust score in half.

Overwatch

Players can be reported to Overwatch. Moderators review recent games and statistical flags, then dismiss the report or convict. Conviction is not automated — flags are starting points for human review, not proof.

Community rep

Signed-in players can give +REP or −REP on profiles they've played against. Rep does not directly change the trust percentage, but it reflects how the community rates someone's sportsmanship and fair play over time.

Important: Trust ratings are statistical summaries. A high score does not guarantee fair play; a low score is not a ban. Use them as context alongside rating, game history, and your own judgment.

Look up a player

Search by name or username, or browse recently finished games.