Nobody stakes money on a game they believe is rigged. So before ChessBit could be a real-money chess platform, it had to be a fair-play platform — and that system has a name: Argus, after the hundred-eyed watcher of Greek myth. Here's how it's designed, told as plainly as we can without handing cheaters the manual.
Engine assistance is the obvious threat: a phone in the other room whispering moves. But a detection system that just asks "did they play like Stockfish?" fails twice over — strong humans do find engine moves in forcing positions, and smart cheaters only consult the engine at critical moments. Single-signal detection either misses the careful cheater or flags the brilliant honest one. Both failures are fatal on a platform where a flag touches someone's money.
Argus is built as independent layers of evidence, and no single one acts alone:
- Engine analysis. Every staked game is analyzed move by move after it ends — against best play, and against what a player at that rating should plausibly find. Brilliance in one game is chess; sustained brilliance beyond your level, precisely at the moments that matter, is a pattern.
- Statistical models. Accuracy, critical-move performance and timing, measured against your own history and rating band — a profile a cheater has to beat consistently, not once.
- Behavioral and device signals. Patterns across games, sessions, accounts and devices — the layer that catches account-sharing, sandbagging and collusion, which engine analysis alone never sees.
- Stake-aware weighting. The more money on the line, the more scrutiny the game gets. Testing the waters at low stakes doesn't build a "clean record" that launders a high-stakes heist later.
A single suspicious game is never enough on its own. Enforcement wants multiple independent signals pointing the same way, across a pattern of games — and ambiguous cases route to a human before anything stands.
The design insight that makes all of this enforceable: payouts wait for clearance. Winnings from every staked game sit in escrow until Argus clears the game — normally a formality you won't notice. There's no chasing money after it left; on a crypto platform, "clawback after the fact" is a fiction, so the check happens before the payout, every time.
The same gate holds tournament prize pools until fair-play review completes. We'd rather pay a day late than pay a cheater.
No detection system is perfect, so the process assumes fallibility:
- Every enforcement carries a 30-day appeal, reviewed by a different reviewer than the original decision, with an evidence summary available on request.
- A successful appeal restores everything — account, balance, record.
- Funds are only ever forfeited after the appeal window closes, with human sign-off. Argus flags; it doesn't confiscate.
And the honest opponent never pays for our mistakes: compensation already paid out stays paid.
When a game is voided for cheating, the honest player gets their full stake back plus a bonus equal to the fee the cheater paid. ChessBit refunds its own fee too — we keep nothing on a voided game, so we lose money every time a cheater plays. That's deliberate: it makes fair play a financial instinct for us, not a PR line.
For the cheater: dishonest winnings clawed back, escalating suspensions up to a permanent ban, frozen funds distributed to documented victims after the appeal window — and the identity behind the account (documents, payment methods, devices) denylisted, so a ban isn't an inconvenience you fix with a new email.
Thresholds, weights, and exactly which moments trigger which checks stay unpublished — publishing them would be a cheater's instruction set. The principle is public, the curve is not.
If you suspect an opponent, report privately — in-app or via a Discord ticket — and let the system look. Public accusations help nobody and hurt honest players; evidence-based review is the whole point.
Full details in the Fair Play & Argus help article, and the Fair Play Policy inside the app.
