Where does the underlying data come from?
Market data is sourced from exchange order books, on-chain transaction records and publicly available sentiment indicators. No single provider is treated as authoritative; inputs are cross-checked against each other before being passed to the predictive layer.
How are the models trained, and how often are they reviewed?
Models are trained on historical market data covering multiple volatility regimes, not a single bull or bear period, to reduce the risk of overfitting to one type of market behaviour. Parameters are reviewed on a scheduled basis and adjusted when monitored performance drifts from expected ranges.
What are the stated limitations of the system?
No model can eliminate market risk. Guardrails reduce exposure to abrupt drawdowns but cannot prevent losses entirely, particularly during periods of extreme illiquidity when exit prices may differ materially from the last observed price. Past model behaviour is not a reliable indicator of future performance.
Who retains control over capital?
Account connections default to read-level access. Any execution permission is granted explicitly by the user and can be revoked at any time, with position sizing limits set by the user rather than the platform.
How is risk tolerance configured?
Each user sets a maximum allocation and a drawdown limit before any position is opened. These parameters govern every automated action taken on the account and can be amended between sessions, not mid-position.
