Strike runs five systematic equity-options strategies — each targeting a specific volatility edge — under a closed-loop agent stack that calibrates parameters per regime, allocates risk across strategies, and learns from every fill.
Most option books are either delta-directional with a vol assumption bolted on, or vol-only books that ignore second-order effects (vanna, charm, dispersion co-movement). Strike treats vol as the primary surface and trades equity options against it — agent by agent, regime by regime.
Every trade starts from a vol view — surface dislocation, dispersion mispricing, or event-window setup — then expressed in the cheapest options structure (vertical, calendar, dispersion basket) given theta, slippage, and risk constraints.
The meta-agent will not size into a vol regime it has not seen in walk-forward. Each sizing change requires fresh out-of-sample evidence, not a calendar — that’s how the closed loop stays honest.
Each strategy targets a specific structural edge in equity options. On its own, each is a systematic rule-set. The agent layer is what makes the stack adaptive — it tunes parameters per regime, allocates risk across strategies, generates post-mortems on every fill, and gates each step on out-of-sample evidence.
Sells the over-priced body and buys the wings of the implied vol smile when the convexity premium clears transaction costs. Captures structural butterfly mispricing on cash-settled indices.
Agent edge: the credit-vs-cost threshold and signal pruning are tuned per session from the live post-mortem distribution rather than left as static config.
Scalps less-liquid stock options on lag, theoretical mispricing, order-flow imbalance, and passive maker quotes — anchored on a vega-weighted Black-Scholes IV calibrated continuously across strikes.
Agent edge: per-name IV priors, signal decay rates, and maker fill probabilities are recalibrated from realised fills, not hard-coded.
Per-stock smile fit identifies strikes whose smoothed residual is an outlier vs the cross-strike scale; trades vega-neutral pairs — short the over-priced strike, long the under-priced.
Agent edge: when post-mortem residuals stop mean-reverting on a name, the agent flags a smile regime change and pauses the name until the surface stabilises.
Compares Garman-Klass realised vol to ATM straddle implied; sells the straddle when implied trades persistently rich and delta-hedges with futures. Exits on convergence, delivery risk, or stop.
Agent edge: threshold and persistence window tune per underlying from its rolling RV/IV behaviour; the meta-agent steps the strategy out of names where the gap has compressed.
Schedule-driven short straddles around named events — RBI policy, FOMC, Budget, large-cap earnings — entered at announcement and exited as the post-event IV crush plays out.
Agent edge: per-event memory of historical crush magnitude drives entry size and exit target, instead of one fixed crush assumption across all events.
Four specialist agents wrap the five strategies — research, signal, execution, risk — and a meta-agent sits above them, allocating risk across strategies and regimes. Every decision is logged, every fill replayed, every parameter recalibrated. That feedback loop is what turns a static rule-set into an adaptive book.
Maintain the inputs each strategy depends on: per-name vol surfaces, RV/IV ratios, smile residuals, event calendars, and screener-tier universes. Re-fit on every scan; flag regime changes before signals are produced.
Translate each strategy’s edge — convexity premium, IV mispricing, smile residual, RV/IV gap, event crush — into a sized trade. Walk-forward tested per regime; signals below threshold are dropped, not traded smaller.
Build the legs each strategy needs (butterflies, single-name calls/puts, vega-neutral pairs, ATM straddles, futures hedges) with explicit slippage, theta, and delivery-risk budgets baked in.
Enforce per-leg, per-strategy, and book-level limits — Greek exposures, hold time, event windows, physical-settlement gates. Veto authority over the meta-agent. A trade is killed before it’s sized if any limit would breach.
Sits above the four specialist agents and the five strategies. Classifies the current vol regime (calm / clustered / crisis), allocates risk budget across the strategies, resolves conflicting signals, and writes every decision to an audit log replayed against paper fills (and, eventually, live fills). No allocation change without walk-forward evidence.
Sharpe, drawdown, hit rate, and convergence statistics will appear here once we have a meaningful out-of-sample window. Until then, this panel stays deliberately empty rather than padded with backtest-only numbers.
Disclaimer: Past performance, simulated or live, is not indicative of future results. Nothing on this page is investment advice or an offer to sell securities.
Our other product. A daily synthetic-NAV engine that re-marks private-market portfolios using transaction panels, web traffic, headcount, and geospatial signals. Live demo today — best backtest cohort R² 0.76, ~50 days lead vs the GP report cycle.
Drop your email below to be notified when Strike exits internal testing and opens to a small group of early partners. No marketing emails — only product status updates. While you wait, our other product, AltVal, is live in demo today.
