It asks one question of every closed trade: is this robot losing money for longer than chance explains? It shows your live numbers beside the backtest's, each with its margin of error. It says nothing until it has enough trades to judge. It does not tell you the cause. It shows what changed and what to check first.
How does it know my robot has stopped working, and how fast?
A robot that earns less than its backtest but still makes money gets no alarm. You see the gap between live and the backtest, and the call is yours. The alarm is for a robot that is losing.
20 trades and 14 days is when Oathra starts to speak, not when it alarms. It raises the alarm only on strong evidence: on a healthy strategy, about one false alarm in a hundred.
Five causes are common: live costs wider than the test, a result that was luck, a market that changed, a different data feed, and slippage. Check your costs first.
Too early means there are not enough trades yet, and Oathra will tell you when there are. Can't judge means trades that close in pieces or do not add up, so Oathra gives no number.
A losing stretch the backtest also had is normal. Oathra asks one question of every closed trade: is this robot losing money for longer than chance explains?
Oathra does not tell you what to do. Only you can decide whether to stand the robot down. Check your costs first.
A tester shows the promise. A tracker shows the result. Oathra checks one against the other and tells you when to worry.
The journal is one tool in Oathra. The centre is the test and the live tracking.
Tracking needs 20 trades and 14 days before its first reading. A robot linked on the trial's first day reaches that on the last day at the earliest. A robot that already has that much history can get one sooner.
You are told when the alarm is raised. The alarm goes by email.