Editorial Policy
OptionPilot is published by AInvest Fintech Inc.. This page sets out who produces what, where every number comes from, what we refuse to publish, and what we have got wrong and corrected.
Last updated . Editorial contact: editorial@ainvest.com.
Who publishes this
OptionPilot is a product of AInvest Fintech Inc., the company behind ainvest.com. It is not an independent publication and does not present itself as one. Corporate information, the Terms of Use, the Privacy Policy and the AI Risk Disclosures are linked in the footer of every page. Reach the desk at editorial@ainvest.com; see Contact.
Which pages a person wrote, and which a program produced
Most of this site is generated, and we say so on the page rather than implying an author. Being explicit about this is the point of the section.
- Generated from the options pipeline, no human author: per-ticker pages (max pain, option chain, IV, gamma levels, expected move, put/call ratio), the ranked hubs, and the daily unusual-activity digests. Each carries the session it priced. No byline is shown because no person wrote it.
- Written and edited by the desk: the strategy guides, the glossary, and the explanatory posts. These carry a desk byline and a date.
- Never: we do not publish a named human byline on a page a human did not write, and we do not attribute machine-generated pages to a person.
Where the numbers come from
- Options and index data — ThetaData, settled end-of-day chains.
- Equity prices and bars — Polygon.io.
- Macro, earnings dates and filings — Financial Modeling Prep.
Open interest is reported by the OCC with a one-session lag. Every page built on open interest therefore describes positioning as of the last settled session, not live intraday flow, and says so. Data pages are recomputed each morning before the open and print the date they priced.
What we refuse to publish
A number that can be computed is not always a number worth printing. Three gates run before anything reaches a page, and each exists because of a specific way the output went wrong:
- A 10% cap on the at-the-money bid–ask spread on any page that ranks implied volatility. Implied vol is solved from the quote midpoint, so a wide spread inflates it — and the error only ever points upward, which means illiquid names sort straight to the top of an ungated list. Ungated, our own “highest IV” ranking returned names quoted 30–60% wide.
- A liquidity floor of 10,000 contracts of chain open interest, a $5 share price and at least 8 listed strikes before a ticker is ranked against others. Premiums quoted on a chain with a $0.05 bid and a $1.40 ask are arithmetic, not prices.
- A suspect flag instead of a silent number. Where a 52-week IV range contains an outlier, the derived rank disagrees with the percentile. We mark those values and exclude them from rankings rather than publish them.
Corrections
We correct in place, say what was wrong, and keep the record. Write to corrections@ainvest.com and we will answer. Two material corrections:
Our expiry-curve builder applied the Brenner–Subrahmanyam approximation for a single option to a straddle, which is two options. The published figure was out by a factor of exactly 2.0000 on all fifteen live expiries, for months. It was caught by a reader comparing our forward IVs against another provider’s term structure — not by an internal alert. The formula was fixed and the affected pages recomputed.
Corrupted at-the-money IV readings over a two-week window set an impossible 52-week high, which collapsed the derived rank for thousands of names — a page could report “IV rank 4, excellent for buying volatility” against a 52-week high the name had never approached. The distribution was the tell: a per-name percentile pooled across names should span 0–100 and ours had a median of 9.3. The quality gate had been counting rows without ever looking at a value. We now rank on the outlier-robust percentile and flag disagreeing ranks rather than publishing them.
What our scoring does not do
OptionPilot scores every optionable US name across five pillars. We tested whether those scores predict returns, and published the result even though it is unflattering.
Backtested over 172 trading sessions (17 December 2025 to 4 September 2026), roughly 2,943 liquid names per session and about 500,000 ticker-days. The cross-sectional information coefficient against forward returns was +0.0001 at one day, with a 95% confidence interval of [−0.011, +0.011]. A tradeable factor runs 0.02–0.05, so the interval sits entirely below it. Deciles are flat, with no gradient.
This is an upper bound that rules an edge out — not a failure to detect one. We therefore do not claim, and you should not read, any of our scores as a prediction of future returns. They are a screen over observable conditions: how expensive options are against a name’s own history, where open interest sits, how liquid the chain is, what is priced into the next event.
Not investment advice
Nothing here is a recommendation to buy or sell any security, and we hold no positions on your behalf. Options involve substantial risk of loss and are not suitable for every investor. See the Risk Disclosures linked in the footer.
Where we use AI, and where we do not
The ranked pages and digests are produced by our own quantitative pipeline from vendor market data — arithmetic and statistics, not a language model. Where a language model assists in drafting explanatory prose, a person at the desk edits and is accountable for it, and no figure in a published page originates from a language model: every number traces to the pipeline and the vendor data behind it.
Email corrections@ainvest.com with the URL. Both corrections above came from outside the company; we would rather hear it.