How do you backtest options flow?
By Faizan Ahmad. Updated .
Describe the rule in plain words in the Tradytics backtester and it runs on every options flow order since March 31, 2025, 19.96 million orders over 404 sessions, with each option filled at the real bid and ask at that moment and $0.65 a contract, then shows the equity curve against SPY, every trade, Sharpe, drawdown, win rate and the code it ran. No coding, and no option priced from a model.
The steps
- Click Backtest beside Chat on any page, or open tradytics.com/backtest.
- Write the rule the way you'd say it: "buy the contract when the same call is bought at the ask three times in one session, hold to expiry", or "the ten stocks with the most call premium bought at the ask each week, held for the week".
- Read the result: the curve against SPY and against buy and hold, every fill and round trip, the same rule after trading costs, and on a final stretch of the period left untouched while the rule was built.
- Change it in the same thread ("only opening orders", "use the 30-delta call instead") and the analyst reuses what it built.
What the flow history holds
| Data | Span |
|---|---|
| Options flow orders: side, premium, sweeps, blocks, opening orders, size against open interest | Every session since March 31, 2025 |
| Option quotes for fills | Real bid and ask since March 2022 |
| Stocks and ETFs | Daily since 2003, split adjusted, dividends paid; minute bars |
| Dark pool prints, to join with flow | Since October 2024 |
| Kalshi and Polymarket markets | Since 2021 |
The orders are the full OPRA tape stitched into orders and tagged bought or sold, the same flow as the live feed, so a rule tested here reads the tape you trade from.
What a 560 million combination flow study learned
The study behind Tradytics Signals tested 560 million pattern and horizon combinations on 10.9 million options prints over 378 sessions. Most of what it threw out failed for reasons any flow backtest has to handle:
- Fill on the option, not the stock. An overnight pattern in illiquid call buying was worth 69 bps in the stock and lost 37% per trade on real option quotes: 6% of trades won against a median spread of 18%. Stock moves of 10 to 70 bps don't cover a spread plus theta.
- Count once per ticker per session. Repeated prints in one ticker on one day inflated early results; counted once, the intraday effect shrank to 2 to 5 bps.
- Per-ticker direction doesn't carry. A ticker's flow direction matched its next window 49 to 51% of the time.
- Add dividends back. 14 of the first 15 survivors were dividend capture: deep in-the-money call volume the day before the ex-date, then the stock "falling" by its dividend.
- Picking recent winners is noise. Trading the patterns that worked over the trailing 60 to 90 sessions made 1.6 bps, no better than random picks.
How the backtester handles them
- Options fill at the ask to buy and the bid to sell, from real quotes, with $0.65 a contract; a contract held to expiry pays its intrinsic value.
- A rule only sees what was known at the time: flow orders after they printed, fills after the signal.
- Prices are split adjusted with dividends paid, so ex-dates don't fake a signal.
- Every result is checked against trading costs, buy and hold and the untouched final stretch.
Plans
Free runs 1 backtest a day, Plus 5 ($39 a month) and Pro 25 ($69 a month). To run your own tests on the raw orders, Pro includes the options flow API with every session since March 31, 2025.
Common questions
Can I backtest unusual options activity?
Yes. Describe what counts as unusual (size against open interest, repeat buying, opening orders) and Tradytics tests it on every flow order since March 31, 2025.
Does the backtest use real option prices?
Yes. Every option fills at the real bid and ask at that moment, from March 2022 on, with $0.65 a contract.
Do I need to code?
No. You write the rule in plain words; the analyst writes and runs the code and shows it with every trade.
Why did a flow pattern that worked on the stock lose on the option?
Spread and time decay. In the Tradytics flow study, a 69 bps stock pattern lost 37% per trade once filled on real option quotes.
How far back does the flow go?
Every session since March 31, 2025: 19.96 million orders over 404 sessions.
- Most bullish options flow on 2026-11-05: META ($92.31M net), LULU ($57.29M net), NOW ($38.58M net), STX ($23.30M net), BE ($20.91M net).
- Most bearish options flow on 2026-11-05: MRVL ($195.81M net), AMD ($121.12M net), TSM ($107.75M net), PFE ($89.87M net), BX ($68.45M net).
Updated 2026-11-06 07:48 UTC