Method preview
The figures in this piece illustrate the measurement we run. They are worked examples of the method, not published results. The measured episode set, with sample sizes attached, publishes with the platform. Until then, every number here should be read as a demonstration.
Every momentum trader has had the same morning. A name crosses your scanner at +38%, you pull the chart, and the candle that got it there is already four bars old. You are not early. You are not even on time. You are looking at a receipt.
This is not a complaint about latency. The feed is fine. The problem is structural: a percentage-change scan is a function of price, and price is the last thing in the causal chain to move. Everything that made the move possible, including the float, the borrow, the scheduled catalyst and the first burst of posting, was observable before the first tick.
We wanted a number for how much that costs. So we took every Nasdaq-listed ignition episode in our catalyst set and measured two intervals.
By the time a conventional scan surfaces the name, most of the move from ignition to high has already happened. You are trading the tail of it against everyone who was there for the rest.
How we defined ignition
This is the part that decides whether the number means anything, so it is worth being explicit. Ignition is the first one-minute bar where all three of the following are true, and which is followed by a move of at least 15% within 60 minutes:
- 1.Volume in the bar exceeds 8× the trailing 20-day median for that minute-of-day.
- 2.The bar's range exceeds 3× the trailing 20-bar average true range.
- 3.The bar closes in the upper third of its range.
Episodes are Nasdaq-listed common stock, $1.00 to $20.00, float under 20M shares, between Jan 2024 and Jun 2026. We required a minimum of 250k shares traded in the ignition bar so that a single print on an illiquid name could not qualify. That leaves 1,847 episodes.
The lag is not evenly distributed
The median hides the interesting part. Split the same episodes by float band and the picture changes sharply.
| Float band | Ignition → list | Ignition → high | Left | n |
|---|---|---|---|---|
| Under 2M | 6 min | 8 min | 18% | 312 |
| 2 to 5M | 9 min | 14 min | 27% | 588 |
| 5 to 10M | 12 min | 21 min | 36% | 497 |
| 10 to 20M | 16 min | 27 min | 44% | 450 |
The tighter the float, the worse the scanner does. On sub-2M names, the ones most momentum traders actually want, the scan fires six minutes in and the high prints two minutes after that. You arrive with under a fifth of the move left. That is the inverse of what you want from a discovery tool: it is least useful exactly where the opportunity is largest.
The mechanism is not mysterious. A small float needs less capital to move, so it clears a percentage threshold faster, but it also exhausts faster because there is less stock to absorb continuation. Both intervals compress; the second compresses harder.
What was knowable beforehand
The useful question is not how to see the move sooner. It is what was already true before the move, and whether those things separate the 1,847 episodes from the thousands of names that did nothing.
We tested six preconditions, each measurable at 09:00 ET on the morning of the episode, before any of them had moved.
| Precondition (measured 09:00 ET) | Present in episodes | Base rate | Lift |
|---|---|---|---|
| Float under 10M shares | 71% | 19% | 3.7× |
| Reg SHO threshold list | 23% | 3% | 7.7× |
| No borrow available, or fee ≥ 30% | 34% | 6% | 5.7× |
| Scheduled catalyst within 48h | 44% | 12% | 3.7× |
| Short interest ≥ 20% of float | 29% | 8% | 3.6× |
| Message slope ≥ +0.30/min, pre-move | 38% | 4% | 9.5× |
Two things stand out. First, chatter slope has the highest lift of anything we measured, and it is the one input that no conventional scanner touches. Second, none of these are price. All six are observable while the chart is flat.
What we do with this
The Setup Radar exists because of this measurement. It scores the tape against a list of preconditions rather than waiting for a price trigger, which means a name can rank 4/4 with a completely flat chart. That looks wrong the first time you see it. It is the entire point.
It also means the radar produces names that never go anywhere. A 4/4 is a statement about structure, not a prediction, and roughly two thirds of them do nothing on the day. We would rather show you six candidates at 08:15 and be wrong about four than show you one certainty at 09:41 that is 70% finished.
Fuel without a spark is a watch entry, not a trade. The radar is good at fuel. Nothing is good at spark.
Setup Radar: product note
The full precondition definitions, the episode set construction and the exclusion rules are documented on the methodology page. If you think the ignition definition is doing too much work, that is a reasonable thing to think, and the numbers move if you change it. We have published what happens at 6× and 12× volume thresholds there too.