Method

The workflow has four stages. A view is stated as a direction, an expectation for implied volatility, a horizon, and whether uncapped loss is acceptable. Those choices are compared against the documented profile of each template. A chosen template then builds standard strikes around spot, snapping to listed contracts when market data is loaded. Finally the position is stressed: payoff, Greeks, a price-and-time table, model-implied probability and exact repricing under a scenario.

Dragging a strike changes the legs and recalculates every dependent output. Editing any leg renames the position to Custom position, so the interface never implies that a modified structure is still the untouched textbook template.

Commercial tools described as optimisers search a large option chain against a target price, a volatility forecast, a budget or a stated objective. None of those four are available here, so describing the result as optimisation would claim more than the inputs support. The tool matches structures instead.

Each template carries three documented labels. These are its usual directional exposure, its usual volatility exposure, and whether maximum loss is defined. A user selection is compared against them and reported as matches, adjacent or conflicts. Matching templates sort first; ties keep the published order, which makes the ranking deterministic.

These labels are conventional starting points rather than guarantees. Once contracts are selected, the position's actual net Greeks are the source of truth, and volatility exposure in particular moves with spot, strike and time. The horizon sets the template's initial days to expiry and is not a fourth ranking factor; treating it as one would imply that 30 days is better or worse than 60 without any forecast to justify it.

No expected return, drift, expected value or capital budget enters the ranking. Beyond creating false precision, such inputs tend to count the same market view more than once.

Practitioner material sorts option structures on two axes before anything else, and the templates follow that convention.

The first is cash flow. A debit is paid to open, so maximum loss is the premium and the position needs the underlying to move enough to justify that cost. A credit is received, so maximum profit is capped at that credit and the position is generally trying to be right about a range rather than a direction.

The second is how far the underlying must travel. A one-standard-deviation range over the horizon, approximately S·σ√T, separates moves that stay inside the expected range from those that break out of it. That distinction is what separates a bull call spread, which profits from an ordinary move, from a long call, which usually needs an outsized one.

The loss column asserts one specific thing: whether loss is unbounded. A cash-secured put is marked defined because the underlying cannot fall below zero, so its worst case is the strike less the credit received, a large loss but a finite one. Only structures containing a naked short call are marked uncapped.

Each option leg is priced independently with Black–Scholes–Merton, then multiplied by its signed quantity and the 100-share contract multiplier. Stock legs are valued directly. Calls and puts take spot S, strike K, time T, risk-free rate r, dividend yield q and the leg's implied volatility σ.

The initial model premium is the sum of the unshocked model values: positive is a debit, negative a credit. P&L is future model value minus that initial value, which keeps one stable baseline rather than mixing theoretical values with quotes that move. It is not a claimed execution price.

Price, time and volatility controls re-run the complete model rather than approximating it. Greek attribution then compares the exact change in model value against the local Taylor estimate Δ·dS + ½Γ·dS² + Θ·dt + ν·dσ. The residual captures higher-order and interaction effects, and it grows as the shock gets larger, which is why it is reported alongside the estimate rather than absorbed into it.

Market data comes from Yahoo Finance through yahoo-finance2, supplying the snapshot, expiries and option chain. Spot, rates, dividends and at-the-money implied volatility can initialise the model, and a selected listed contract can supply a leg's implied volatility. Bid, ask, mid, open interest and volume stay marked as observations beside the model rather than inside it.

Yahoo's own delay table lists US options data from OPRA as delayed by fifteen minutes. The interface therefore says “Yahoo” or “market data” and never “live”, and treats every quote as informational rather than executable.

Automated tests compare prices against textbook references, enforce put–call parity, check Greeks against finite differences, verify standard payoff identities and tail behaviour, partition multi-breakeven probability regions, and confirm that matching is deterministic. Template labels are also tested against the payoff engine, so a structure cannot claim defined loss unless the engine agrees its loss is bounded.

These checks validate implementation consistency. They cannot validate the model assumptions themselves. Those are addressed in section 8.

The first three items are design limitations of the matcher rather than of the pricing engine, and are the most likely to change.