Equity backtesting asks “does this rule make money?”. Options backtesting asks that plus “was the option priced correctly, which contract should have been chosen, and what did the greeks do on the way?” Each extra question is a place where a test can quietly flatter itself.
The three variables that break naive tests
- ▸Valuation model. An American-style contract can be exercised early, a Heston surface has a volatility smile, and a plain Black-Scholes number assumes neither. Whichever model your test uses, it should be named — and the same strategy priced under two models should be expected to disagree.
- ▸Contract selection. A backtest that picks the at-the-money strike nearest the signal is making a rule you never wrote. Selection by moneyness, DTE, or delta changes results as much as the entry logic does, so it belongs in the strategy, visible and testable.
- ▸Fill realism. Options spreads are wide and liquidity thins fast away from the front month. A test that fills at mid-price on a three-contract order is measuring a market that does not exist.
What to demand from the workflow
The strongest pattern is the same one that works for equities: screen broadly with a fast engine, then re-run survivors with full mechanics. In options that means using vectorized tests to discard most ideas cheaply, then confirming the remainder order-by-order with real greeks along the path, explicit costs, and a validation step — walk-forward and Monte Carlo — that reports the overfitting risk before any capital is offered. Paper trading the validated artifact closes the loop: the strategy you tested is the strategy that runs.