Backstop is an autonomous paper-trading system built around a simple inverse-volatility portfolio strategy and a policy firewall that governs every order before it reaches Alpaca. The strategy allocates risk across an eleven-asset basket, rebalances only when weight drift exceeds its configured threshold, and keeps a cash buffer. The firewall independently enforces account identity, position and notional limits, market-data freshness, pending-order exposure, drawdown controls, and unsupported-order-shape blocks. Broker positions and open orders are reconciled through authoritative snapshots; unavailable or inconsistent exposure data fails closed instead of being guessed. A session drawdown killswitch blocks risk-adding orders after the configured loss threshold, while a narrow deleveraging exception may permit only a provably exposure-reducing equity sell. Reactive option hedges remain proposal-and-audit only, and option execution is currently disabled. Every policy decision is written to a tamper-evident audit chain and surfaced through a human-readable dashboard. The project was tested adversarially across four audit rounds and executed against a pinned $100,000 Alpaca paper account. Alpaca records verify ten filled orders, nine resulting positions, and zero open orders. Backstop deliberately distinguishes what is automated, what is merely proposed, and what remains a disclosed limitation.
Category tags: