🤓 Latest Submissions
VRP Engine: Autonomous Options Agent on Alpaca
VRP Engine is an autonomous options agent built for Alpaca's Options Alpha Agents track. It trades only on Alpaca paper and targets the variance risk premium: the gap between the volatility the option market implies and the volatility the underlying actually delivers. Each cycle the engine measures realised vs implied volatility across a multi-underlying universe, reads trend and IV term slope, then proposes defined-risk structures—credit spreads, debit spreads, or iron condors—never naked shorts. A trade opens only when a positive probability wedge says the specific structure is mispriced, not merely that IV looks rich. Sizing starts from fractional Kelly on modelled edge and is clipped by per-trade, per-underlying, correlation, buying-power, and liquidity budgets. Risk is hard-coded, not prompted. Every proposal passes review_proposal before any order: defined-risk proof from the legs, portfolio stress, account breakers, and trading-window checks. An optional LLM analyst briefs the regime and may soft-veto; it cannot bypass budgets. Dry-run is the default; live paper execution is explicit. Alpaca surfaces have distinct jobs: alpaca-py submits multi-leg tickets, the CLI verifies the book before and after every order, and an MCP client researches quotes and regime context read-only. A Streamlit dashboard and JSONL journal make decisions, risk, and paper P&L inspectable for the full competition week—identify, decide, size, risk-check, manage, and reconcile in one loop.
4 Sep 2026
