How Signals Are Graded

The rule was fixed before the record started and does not move. This page is the whole of it: what a Signal is, what a win is, what is excluded, and how versions are reported.

What a Signal is

A Signal is a single institutional options print — one contract, one side, one fill — that clears a fixed bar: at least $100K in premium and a score of 60 or higher on a 0–100 scale. The score is computed once, at fire time, from the inputs below. It is a measure of how far outside the normal tape the print sits and how well-supported it is; it is not a probability of profit.

The fill used for grading is the volume-weighted price printed on the alert. No better fill is assumed after the fact.

Scoring inputs

Weights are versioned and not published; the categories are. Every Signal's detail view inside the platform shows which of these contributed and which did not.

• Premium tier — how much money the print represents.
• Aggressor and moneyness — whether the print lifted the ask or hit the bid, and how far the strike sits from spot.
• Sweep and opening interest — multi-exchange sweeps, and volume that exceeds the contract's existing open interest.
• Repeat hits — the same contract printing again in the same direction.
• Runway to expiry — days left on the contract relative to the print's size.
• Ticker consensus — other qualifying prints on the same name leaning the same way.
• Earnings runway — whether an earnings date falls inside the contract's life.
• Dealer-positioning confluence — where the strike sits against the day's gamma map: with a wall, at a magnet, or against structure.

Why first touch decides

A contract can print both rails inside its life. Grading on the peak alone would count every one of those as a win; grading on the close would ignore what price actually did in between. First touch answers one narrow question — which rail did price reach first — and it is checked bar by bar from one-minute trade data. When both rails print inside the same bar, the order is unknowable, so the tie goes against the signal.

Why peak is not a return

Peak % is the best print the contract reached after the alert. It is published because it describes the shape of a signal's life, and it is always shown next to the worst dip that came before it. A peak is a print, not a fill anyone is assumed to have received, so it is not money made and is never presented as one. The record contains no dollar results, no member outcomes, and no one's trades.

The no-edge baseline, and why the matured cohort leads

The two rails are not the same distance from entry. The stop (−50%) sits 2× as far as the target (+25%), so a contract with no directional edge at all touches the target first about 66.7% of the time (50 ÷ (25 + 50), the driftless first-passage rate; with time decay and expiry-as-loss it lands within a couple of points of that). A raw win rate near that number is what the yardstick alone produces. Every rate on the record page is therefore shown next to this baseline as edge in points, with the 95% interval read against the same line: when the interval includes the baseline the rate is not distinguishable from no-edge, and the page says so. The baseline is computed from the rails, so if a rail ever moves it moves with it and the change is disclosed above.

The record page leads with the matured cohort: signals whose expiry has passed, or that fired 14 or more days ago. Wins resolve in under a day on average because the target is the nearer rail; losses take weeks because most of them are scored at expiry. While a cohort is still open its graded set holds more of its eventual wins than its eventual losses, so the all-graded number runs hot. Both numbers are published with their own n and interval.

What is excluded, and still counted

• No verifiable fill — the print carried no usable price.
• No timestamp — the print could not be placed on the tape.
• Multi-leg structures — spreads and combos, until per-leg pricing exists.

Excluded signals stay in the total and are listed by reason on the record page. They are not deleted, and they never move into the win or loss columns.

Versioning

Scoring weights carry an engine version. When weights change, a new version begins and the record is reported per version (current: v2). Signals graded under an earlier version are never re-scored under a later one, and the grading rule above is the same across all versions.

Rates always ship with their sample size and a 95% Wilson interval. Any slice with fewer than 30 graded signals is shown in amber and is not ranked against other slices.

Questions

What counts as a win?

A Signal is a WIN when the option's price prints +25% above the alerted fill before it prints −50% below it. Whichever rail is touched first decides; a tie inside the same bar counts as a loss. If neither rail is touched by expiry, it is a LOSS.

Is the win rate a return?

No. It is the share of graded signals whose price reached the upper rail first under a fixed measurement rule. Peak % is a high-water mark, not money made, and the page never reports dollar results.

Why is a signal MATURING for so long?

The clock runs to the contract's expiry, so a signal on a 45-day option can stay open for weeks. Maturing signals are counted and shown, but excluded from every rate until they resolve.

Why are index and 0DTE listed apart?

Same-day and index contracts move differently from single-name dated options, and blending them would flatter or bury the headline. Each gets its own numbers, with its own sample size and interval.

What does the 95% interval mean?

It is a Wilson score interval around the win rate. With few graded signals it is wide, which is the honest picture; slices under n=30 are shown in amber and never ranked.

Why is the win rate compared to 66.7%?

That is the rate a contract with no edge prints under this rule. The stop rail sits 2× as far from entry as the target, so a random walk reaches +25% before −50% about 50 ÷ (25 + 50) of the time. Edge is the win rate minus that line, in points; a win rate at the line is no evidence of edge.

Why does the matured cohort lead the page?

A win resolves in under a day on average and a loss takes weeks, so at any moment the open signals hold more future losses than future wins. Signals whose expiry has passed, or that fired 14+ days ago, have had time to resolve either way; that cohort is the cleaner read, and the all-graded number is shown next to it.

Can a losing signal be removed later?

No. Every signal that fires is graded and stays in the tape, including losses and exclusions. Excluded signals are listed by reason rather than dropped.

What happens when the scoring changes?

Weights are versioned. A change starts a new engine version, and the record is reported per version so an old cohort is never re-scored under new rules.

Is this a recommendation to buy or sell anything?

No. Signals measure institutional order flow against a fixed grading rule and are shown for information; they are not investment advice, and past results do not predict future outcomes.

Signals are a measurement of institutional order flow graded by a fixed rule; they are not investment advice, and past results do not predict future outcomes. Options involve significant risk.