Introducing

FTO-1

An outcome model built to find the opportunities worth a closer look.

For every single-option contract the Optimizer finds, FTO-1 estimates the chance it reaches a +100% profit target before a −50% stop or expiry.

Try DeriveAI free

It gives you odds, not a score.
The number you see is a calibrated probability. If FTO-1 shows 34%, it means an estimated 34% chance of reaching the stated target before the stop or expiry.
It is deliberately selective.
FTO-1 is most useful when it sees a setup that stands well above the rest. And when the market regime cannot support an honest number, it shows a probability range instead of a false precision: roughly half of all contracts get an exact number, the rest get honesty.
It compares the contracts in front of you.
The model does not judge a contract in isolation. It also asks how it stacks up against the other choices in the same search, then keeps the probability on screen easy to read.
It only looks at what you could know before buying.
The test starts after the trade is opened. Nothing that happened later is allowed to help the model make the earlier decision.
It knows what it doesn't know.
FTO-1's confidence is calibrated. When it's unsure, it says so, instead of pretending.

What you can do now

Know the odds before you commit.
Every recommendation carries a probability. No more guessing whether a 92 score is good.
See what your trades actually did.
Your Scorecard grades every past recommendation: what it did, when it peaked, what holding cost you.
Find out how you trade.
Because every trade is graded, your Scorecard can tell you when your positions typically top out, which market conditions suit you, and what you're leaving on the table.
Know when to sit out.
Some conditions don't have a good trade in them. FTO-1 can tell you that too.

Opportunity Insights

When you run an optimization, FTO-1 doesn't just score the handful of contracts it shows you. It scores every contract available for that stock, often hundreds, and compares the two. A note at the top of your results tells you what it found.

If one of your settings is in the way, it says which one.
Sometimes the strongest contract out there costs more than your limit, or expires outside the timeframe you chose. When that happens, the note says so in plain words, and one tap re-runs the same optimization with that setting changed.
And when there's simply nothing good out there, it says that too.
Some days no contract on a stock is worth taking. FTO-1 tells you the whole board is weak instead of dressing up a weak pick, so sitting out is a decision, not a guess.

What that means for a trader

The upgraded FTO-1 was validated walk-forward on more than 1.2 million historical option contracts across 12 months, every month scored by a model that had never seen it. These are historical modeled results before execution costs, not yet confirmed on a fresh forward month.

0.737

contract-selection quality (ROC-AUC)

Candidate-level, walk-forward validated, up from 0.688 for the prior model. It ranks contracts by pricing not just where a contract can go but the path it has to survive to get there.

2.9%

average calibration error on displayed claims

When FTO-1 shows a probability, it means it: displayed probabilities averaged within 2.9% of realized outcomes, and 4.2% on the unseen second half of the span. Every displayed band landed within 8 points of what actually happened.

44%

of contracts get an exact number

When the market regime makes an honest number impossible, FTO-1 says so: it shows a probability range instead of a false precision. The model abstains rather than display an unsupported number.

See the full validation numbers

When it states a probability, it delivers

The lighter bars show the probability FTO-1 stated; the solid bars show how often those contracts actually reached their target before the stop or expiry. Across 553,000+ displayed claims in the walk-forward test, every displayed band landed within 7.3 points of reality. These are barrier-hit rates, not profits.

Stated probability

Delivered outcome

~6%

~16%

~25%

~35%

~54%

~64%

~76%

~94%

High confidence, checked against the clock

The old model's weakness was short-dated contracts: it priced where a contract could go, not the path it had to survive to get there, so its most confident short-dated calls delivered far less than promised. The upgraded FTO-1 prices the path too. The bars show how often contracts in its most confident group (a stated chance of 80% or more) actually reached their target, previous model beside upgraded, split by time to expiry. Reaching the target means the profit target was hit before the stop or expiry in the historical test; it is not a profit after costs.

Previous FTO-1

Upgraded FTO-1

Over two weeks to expiry

Three to seven days to expiry

What it learned from

Every number FTO-1 learns from is a real optimizer recommendation graded to resolution. The graded record stands at 311,493 searches and 1,391,820 contract recommendations, and every one carries a resolved outcome: target reached, stopped out, or expired. About 4.5 contracts are graded per search.

1.42M

graded contracts behind FTO-1

FTO-1's base model trained on 1,420,035 graded contract recommendations, and its newest layer learned from 1,322,291 of them, each one paired with the market conditions of its day.

1.39M

recommendations graded to resolution

The full graded record: 1,391,820 contract recommendations, every one carrying a resolved outcome, whether the target was reached, the stop was hit, or the contract expired.

311,493

optimizer searches graded

The shared record behind our models. Nothing is simulated from quotes alone: the models learn from what the Optimizer actually recommended and how each recommendation resolved.

It doesn't predict direction.
It won't tell you whether a stock rises or falls.
It doesn't predict price.
No targets, no forecasts.
It answers one question:
Will this reach +100% before it drops 50% or expires? Does this contract double before the trade plan ends?
Change the trade, change the odds.
The numbers assume you're taking profit at +100% and cutting at −50%. Different plan, different probability, and we always show you which one a number is based on.
Single options only.
FTO-1 scores one contract at a time. Debit and credit spreads are scored by FTO-1.5, which reads the whole position.

See FTO-1.5+

Available now

FTO-1 is live in the Optimizer for every user. Nothing to turn on.
Your Scorecard is already populated: every optimization you've run has been graded and is waiting for you.

For readers who want the exact test

	Test scope	Result
Ranking accuracy (ROC-AUC) How often a target hitter ranks above a non-hitter; 0.50 is chance and 1.00 is perfect. Candidate-level, walk-forward validated.	All 1,207,199 candidates · 12 monthly folds	0.7366
Ranking accuracy, prior model The deployed model this release replaces, on the same test.	All 1,207,199 candidates · 12 monthly folds	0.6880
Confidence penalty (log loss) Penalizes confident predictions that are wrong; lower is better. Prior model: 0.5104.	All 1,207,199 candidates	0.4665
Calibration error on displayed claims Average gap between the probabilities shown and the outcomes observed; lower is better.	553,000+ displayed claims	2.9%
Calibration error, unseen half-span The same measurement restricted to the second half of the validation span, which no tuning decision saw.	Displayed claims, second half	4.2%
Worst displayed band The largest gap between any displayed probability band and what it delivered.	10 probability bands	7.3pp
Numeric coverage The share of contracts shown an exact number. The model abstains rather than display an unsupported number; the rest get a probability range.	All displayed contracts	44.2%
Stated ~6% band Observed barrier-hit rate against the stated probability. Intervals in parentheses are 95% confidence ranges where computed.	250,053 displayed claims	7.9% observed (7.8–8.0%)
Stated ~16% band	97,200 displayed claims	19.9% observed
Stated ~25% band	90,125 displayed claims	24.1% observed (23.8–24.4%)
Stated ~35% band	49,878 displayed claims	29.9% observed
Stated ~54% band	8,656 displayed claims	47.3% observed
Stated ~64% band	13,176 displayed claims	60.0% observed (59.2–60.8%)
Stated ~76% band	12,711 displayed claims	68.9% observed
Stated ~94% band	12,348 displayed claims	88.6% observed (88.0–89.1%)

Walk-forward validation across 12 monthly folds with a 30-day embargo, 1,207,199 contracts; no fold saw its own month. Probabilities are modeled barrier-hit rates (take-profit reached before stop-loss or expiry) before commissions, slippage and execution effects, not a probability of profit. These are historical modeled results, not guaranteed returns, and they have not yet been confirmed on a fresh forward month. Not investment advice; past performance does not guarantee future results.

Try DeriveAI free