Frequently Asked Questions

DeriveAI helps individual traders understand the odds before they trade. It combines plain-language research, contract discovery and probability estimates learned from graded trade outcomes, so you can weigh an opportunity with real context behind it.

You make the decision. DeriveAI helps you understand the odds.

No. DeriveAI does not exist to publish picks or tell you when to enter and exit. It helps you research markets, rank the contracts in front of you by their odds, and see when none of them are worth taking.

The product is designed to improve the information around a decision, not replace the person making it.

Search is the answering layer: ask about markets, securities, contracts, strategies or model output in plain language. Optimizer scores the whole chain and ranks it by odds, or tells you when nothing in it is worth taking. FTO is the probability behind that ranking.

Scorecard grades your past recommendations and shows you the pattern in them. Together they form one loop: search, discover, evaluate, decide, learn.

It is the estimated chance the contract reaches its target before it stops out, learned from graded trade outcomes. A 60% estimate is not a promise about one trade.

The important question is whether probabilities remain meaningful across many outcomes. That is why DeriveAI emphasizes calibration instead of a generic score.

No. You need to be comfortable reading a probability and deciding what to do with it. Nothing in the product asks you to understand ensembles or calibration first.

The research is published if you want it, but it is there to be checked, not to be a prerequisite.

The models are built around listed options and the equities, indices and ETFs underneath them, including contract characteristics like expiry, strike and the volatility context around them.

Coverage is described in more detail on the Coverage page, and it will expand over time as the outcome dataset does.

No. DeriveAI is not a broker or an adviser, it does not execute trades, and nothing it produces is a recommendation to buy or sell a security. Options trading involves substantial risk of loss and is not suitable for every investor.

You are responsible for your own trading decisions. The Legal & risk page sets this out in full.

By checking the evidence rather than taking our word for it. DeriveAI is designed to publish model performance over time, including calibration, model versions, evaluation periods and the outcome history behind the record.

Advanced AI shouldn’t require blind trust. The goal is to make model behavior inspectable.

Calibration asks whether probabilities behave the way they claim to. Across opportunities estimated around 60%, for example, did the realized outcome rate land near that level over time?

Because FTO communicates probability rather than a verdict, calibration is one of the most useful ways to evaluate whether those numbers are informative.

We expect it to, and we plan for it. Our research into model decay found that a model's edge is not a fixed property of it — performance rises, falls, recovers and drifts as market conditions change. Later work treated that edge as a time-varying, partly mean-reverting process rather than a constant.

In practice that means models are versioned, monitored after deployment, and retrained on a schedule. When performance moves, it moves in the published record too; we would rather show a decline than quietly retire the number.

FTO-1 and FTO-1.5 are the current outcome-intelligence models, and they are evaluated separately rather than blended into a single headline figure.

Underneath, some systems use a population of specialist models instead of one, tracking how each has been performing and shifting weight toward the ones expected to do better in the coming environment.

The goal is a continuous public record rather than a selected highlight reel, including periods where model performance weakens.

A track record is only useful if it remains visible when the numbers become less flattering.

Because a score is unfalsifiable. If a contract is rated 8/10 and loses, nothing about the 8 can be checked. A probability can be checked, in aggregate, against what actually happened.

Choosing probability is what makes the rest of the transparency possible.

Free is a working version of the product rather than a preview: the whole loop — Search, Optimizer with FTO-1 probability, spreads and multi-leg, the abstention verdict when nothing is worth taking, and Scorecard on everything you run — with unlimited optimizations.

We charge for depth, not for access to the reason the product exists.

Pro Optimizations, FTO-1.5 on anything you run rather than on spreads alone, the Screener so the scanning runs for you, alerts when something you are watching changes, your full Scorecard history and analytics rather than the recent slice, and early access to new models.

It is the same product, further in. Nothing about why DeriveAI is worth using sits behind it.

From your account settings, at any time. Upgrades apply immediately; downgrades take effect at the start of the next billing period, so you keep what you are paying for until then.

Cancelling does not delete your account. You drop back to Free and keep using it.

It stays yours. Cancelling moves you to Free rather than wiping your history, and you can export your Scorecard data before or after you cancel.

Scorecard is only useful because it accumulates, so destroying it as a retention tactic would defeat the point of building it.

No, unless you choose to share it. Scorecard analyses the decisions you make inside DeriveAI, and what it produces is private by default.

If sharing parts of a trading profile becomes available, it stays opt-in and you control which parts.

Email us and a person will read it. We would rather answer honestly and a little slowly than promise a response time we cannot hold to, so we don't publish one.

Questions about the models and the research are welcome too — those often go straight to the people who built them.