The gold log ends while three signals are still open

Choose a clock for each signal

Imagine ten hypothetical gold alerts with a written rule to follow each for sixty minutes from activation. The first qualifying touch of either the target or the stop closes the observation. Four alerts have reached their targets and three have hit their stops. Trading continues when the export ends. Three alerts are still active, with only ten, twenty and forty minutes of uninterrupted observations respectively. All counts are invented for this example.

Four out of seven describes the target share among the completed cases. Four out of ten describes the fraction already observed to hit a target. To report the fraction that hit the target first within sixty minutes for this particular cohort, we still need the ends of three histories. Its possible final count runs from four to seven, including the possibility of expiry. This range describes missing outcomes in ten records and carries no sampling confidence level.

Keep the time already observed

Give each open record its last continuously observed timestamp and a reason for follow-up ending. A ten-minute record establishes survival without either exit through minute ten. It contributes evidence about that interval. If a feed gap began before the last visible quote, end verified follow-up at the start of the gap unless another record establishes the intervening path.

A Kaplan–Meier estimate can describe time to either exit while incorporating right-censored observations. Its interpretation requires that the censoring process leave a comparable set of signals under observation, conditional on the factors used in the analysis. A feed that drops preferentially during violent moves could break that assumption. I would separate scheduled export cutoffs from feed failures and show their counts before displaying a curve.

Source notes: [1] NIST/SEMATECH — Kaplan–Meier estimation

A stop changes the question

For the question of reaching a target first, a recorded stop is a competing outcome. Under the stated rule that signal has ended and can no longer achieve a target-first result. Removing stopped cases as ordinary censored observations can overstate the cumulative probability of reaching the target. A competing-risks cumulative incidence estimate addresses the probability of each first exit by a chosen horizon, subject to appropriate handling of genuine censoring.

For a small archive I would first publish the observed counts, follow-up durations and possible final range. On receiving the rest of the export, append the later evidence to the same identifiers. That preserves what a reader could have learned at the original cutoff.

Source notes: [2] Southern and colleagues — competing risk comparison