If the Same Stock Keeps Triggering,
Is That a Stronger Signal?
When the same stock triggers a signal several days in a row, it feels like "higher conviction." The real measurement showed the opposite. We explain why, and why it's a problem when computing statistics.
Real Measurement — the More Times It Triggered, the Worse the Result
We tallied, by how many times the same stock had triggered a scan, the rate at which it later touched the target (+15%) at least once. As of 2026-08-14, counting only matured picks (20 trading days elapsed, or target reached, or delisted).
| Times Triggered | Later Target-Hit Rate | Sample | 95% Interval |
|---|---|---|---|
| 1x | 45.0% | 140 | ±8.1%p |
| 2x | 38.7% | 62 | ±11.8%p |
| 3x or more | 38.3% | 115 | ±8.7%p |
The direction is downward, but the confidence intervals overlap (36.9–53.1% for 1x vs. 29.6–47.0% for 3x or more). In other words, this difference hasn't been confirmed by the sample. It's not that "repeated triggers are bad" — all we can currently say is "there's no evidence that repeated triggers are good."
These numbers were 42.7 / 36.8 / 29.4% at first publication in 2026-07. As the sample built up, the drop narrowed from 13.3%p to 6.7%p. It's common for an effect that looks large in an early sample to shrink as the sample grows, so we keep updating the figures instead of freezing them, and don't erase a shrinking effect. Check the latest value yourself at /track or /base-rate.
Why This Runs Against Intuition
① It Keeps Triggering Because It "Hasn't Broken Out"
Most pre-breakout signals detect a "just before breakout" state — things like volatility contraction or tightening below resistance. But if this state doesn't resolve and just persists, the stock triggers the scan again every day.
In other words, a repeat trigger can mean not "the signal got stronger," but "nothing has happened for several days straight." A stock that actually breaks out leaves the condition once it does, so it stops triggering.
② The Character of a Stalled Setup
Staying compressed for a long time doesn't always lead to a big rebound. A stock that just keeps drifting sideways with no volume picking up may simply be a stock nobody's interested in.
The Problem for Statistics — Duplicate Observations
A repeat trigger breaks not just how you read the result, but the statistical calculation itself.
Say one stock triggers the signal every day for 3 weeks. Count that as-is and you get a sample of 15. But in reality, you've counted one single event 15 times.
- The inflated sample size makes the confidence interval come out narrower than it really is.
- That one stock's outcome over-represents the whole statistic.
- The result looks like significance that isn't actually there.
The Fix — Group by Episode
DawnScan treats the same stock triggering again within a set window (21 days) as a single episode and counts it only 1x. If it shows up again after that window passes, that counts as a new episode.
This shrinks the sample size, but it makes the number that remains trustworthy. That's why the sample size on the base-rate proof page is much smaller than the raw trigger count.
In Practice
- Don't read a repeat trigger as conviction — there's no evidence for "it triggered a 3rd time, so it must be certain."
- Watch the volume change alongside it — even the same compression looks different depending on whether volume is starting to pick up.
- Display it neutrally — DawnScan's hit-rate page groups the same stock into one row and shows "triggered N times" as information, not a badge of honor.