Small-Cap Gapper Statistics (2026): High of Day, Gap Fills and Day Two in 671 Gap-Ups

Key numbers

All figures: FennEdge archive, sessions 2026-07-01 to 2026-10-02, as of 2026-10-02. Times are Eastern (ET).

Sample and definitions

Universe. For each session from July 1 to October 2, 2026, the study took every security in the FennEdge Top Gappers archive that met three conditions. It was classified as common stock or a common-stock ADR on that date. It opened at least +20% above the previous close, where gap = regular-session open ÷ previous close − 1. Both the open and the previous close were $0.10 or higher.

No size or volume filter. "Small-cap" is the trader's label for this kind of name. The sample itself has no market-cap rule and no volume minimum. Prices are split-adjusted daily aggregates.

High of day (HOD). The high of day is the highest 1-minute high from 04:00 to 19:59 ET. Premarket and after-hours trading count. If the same high printed in more than one minute, the first minute is used. That tie happened in 78 of 671 observations (11.6%), a study figure.

Size. The archive held 676 qualifying observations. The study excluded 5 where minute and daily data disagreed in price scale. That leaves 671 observations, 461 distinct tickers and 66 sessions, with a median of 9.5 per session. None of the 66 sessions was an early close. The median opening gap was +36.0% (n = 671, study figure).

Sessions by month. July had 22 sessions, August 21, September 21 and October 2. Qualifying gappers per session were 8.9 in July (196 ÷ 22), 12.9 in August (270 ÷ 21) and 9.1 in September (192 ÷ 21). These are FennEdge computations for July to September (n = 658).

Other definitions. Day change = close ÷ previous close − 1. Volume is full-day consolidated volume. Percentages are shares of the stated n, rounded to one decimal. Intervals are 95% Wilson score intervals unless a figure says otherwise.

When the high of day prints

The table splits all 671 observations by the first minute of the high. It is a study figure. Each range includes its start and excludes its end.

The early premarket is lumpy. In the study, 132 highs (19.7% of 671) printed in the 4 o'clock hour, and 37 (5.5% of 671) in the 04:00 minute itself.

The opening print was rarely the high. Only 21 of 671 (3.1%) printed their high in the 09:30 minute, a study figure.

Later highs were the minority. As a FennEdge computation, 200 of 671 (29.8%) printed their high at 10:00 or later. Another computation: 315 of 671 (46.9%) did not print their high in premarket, so a later print exceeded the premarket high. Of those 315, 55 printed it after hours.

When the high landed in the regular session (09:30–15:59, n = 260), its median time was 10:06. That is a study figure for names whose extended-hours high happened to fall in regular hours. It is not the median regular-session high of every gapper.

Opening-range view (FennEdge computation, n = 671). These rows use the final high, so they are hindsight.

• High inside the 5-minute opening range (09:30–09:34): 51 (7.6%).
• High from 09:35 to 15:59: 209 (31.1%). On these days, a later minute traded above every earlier minute.
• High in premarket or after hours: 411 (61.3%).
• On a 1-minute range: high in the 09:30 minute 21 (3.1%); high from 09:31 to 15:59, 239 (35.6%).

Bad-print check (study figure). Dropping the 15 observations whose high exceeded 2.5× both the open and the close left n = 656. The median stayed 09:16, the premarket share was 53.0% and the before-10:00 share was 70.6%.

By gap size

Gap size barely moved the timing in this window. The study does not read differences of a few points between these groups.

Timing by opening gap (study figure). Shares are of each row's n.

Opening range and fills by opening gap (FennEdge computation). "Full fill" means the close was below the previous close. "Half fill" means the close was below the gap midpoint.

The 100%+ group has only 83 observations. Its intervals are wide, so small gaps between rows sit within noise.

By opening price

Lower-priced gappers printed their high earlier in this sample. As a FennEdge computation, 97 of 671 (14.5%) opened below $1.00.

Timing by regular-session open (study figure).

Opening range and fills by open price (FennEdge computation).

The study offers one possible reading: higher-priced gappers may include more news-driven or larger-company moves that keep trading through the day. It did not test that idea.

By volume and dollar volume

The study applied no volume minimum. It also checked the public Top Gappers list's default full-day share-volume filter, and the headline timing barely changed. The study has that breakdown.

Dollar volume shows how uneven liquidity was across the sample. Here it means full-day share volume × close, a FennEdge computation over all 671 observations.

The 90th percentile was about 32× the 10th. Percentiles use linear interpolation, and that ratio depends on the choice.

Treat this as a rough proxy. It is a full-day total. It says nothing about how much traded in the minute you might care about, or what the spread was.

Closing below the open and gap fills

Most gappers in this sample closed below where they opened. The study reports 438 of 671 (65.3%) closing below the regular-session open. As a FennEdge computation, the other 233 of 671 (34.7%) closed at or above the open.

A full gap fill by the close was much rarer. The study prints 9.2% of 671 closing below the previous close. The count, 62, and the interval of 7.3%–11.7% come from FennEdge's computation.

A half fill by the close was more common. As a FennEdge computation, 208 of 671 (31.0%; interval 27.6%–34.6%) closed below the gap midpoint, halfway between the prior close and the open.

These are close-basis fills only. They do not say how often price touched the prior close during the session. Intraday touch rates are a planned follow-up study and have no value yet.

Full fills by month (FennEdge computation). July 26 of 196 (13.3%). August 21 of 270 (7.8%). September 14 of 192 (7.3%). October had 1 of 13, from 2 sessions, and is not reported as a month.

Splits that use the final high or the close. The rows below group days by information known only after the session. They describe two kinds of day, not a way to sort stocks at 09:30.

Part of this is mechanical. A stock that fades from an early peak has an early high by definition. A stock that closes well above its open tends to have a late high.

Large intraday runs (FennEdge computation, hindsight-selected, n = 69). These are days whose high printed in the regular session and reached at least 1.5× the open. Of the 69, 56 (81.2%; interval 70.4%–88.6%) closed below the midpoint between the open and the high. The median close was 35.5% below the high. 16 (23.2%) closed below the open. The median HOD time was 10:54. Without the regular-session condition, 174 observations had a high at least 1.5× the open.

Day two

Day two is the next exchange session for the same ticker. Day ones run from 2026-07-01 to 2026-10-01, because the 6 observations from 2026-10-02 have no following session in the window. That leaves 665 day ones from 457 tickers. All figures here are FennEdge computations.

• Day two opened above day one's close: 208 of 665 (31.3%; interval 27.9%–34.9%).
• Of those 208, 69 (33.2%; interval 27.1%–39.8%) closed below day one's close.
• Of those 208, 116 (55.8%) closed below the day-two open.
• Of those 208, 32 (15.4%) opened with another gap of 20% or more.

The other 457 day ones opened at or below day one's close, or did not trade. The archive sees a day two only when it was itself a gap-up, so this group cannot be split yet.

Day two by how day one ended (FennEdge computation).

The intervals in the first two rows overlap. With these group sizes, the difference between them may be noise.

What these numbers do not tell you

They are not a win rate. A day that closed below its open says nothing about where anyone entered or exited. None of these figures includes entries, stops, fills, spreads or costs.

Hindsight grouping. Splits by the final high, the close or a large run use information you only have after the bell. They describe kinds of day. They do not tell you which kind of day you are in at 09:30.

Selection at 09:30. The sample is defined by the official regular-session open. Stocks that ran in premarket but opened below +20% are not in it.

Survivorship. Security type comes from reference data as of each session, and tickers later delisted, renamed or reverse-split stay in the sample. That reduces survivorship bias. It does not remove every classification error, and prices are split-adjusted, so later adjustments can revise them.

One regime. Everything comes from 66 sessions in one stretch of 2026. Even inside it, the monthly numbers moved (study figure):

As a FennEdge computation, the August-to-September premarket gap was 8.4 points (152 of 270 vs. 92 of 192). Its normal-approximation 95% range was −0.8 to +17.6 points, and the true range is likely wider.

Small sample per session. A median of 9.5 gappers per session is not many. 461 tickers produced 671 observations, and gappers on the same day tend to move together. The intervals here treat observations as independent, so they probably understate the real uncertainty.

Small groups. Groups under about 100 observations, such as the 100%+ gap tier, the under-$1 tier and the large-run group, carry wide intervals.

Extended-hours highs. Premarket and after-hours highs can be set on small volume or by a single off-market print. The study removed 5 clear price-scale errors and ran the 2.5× check above, but a few odd prints may remain.

How to use this as a base rate

A base rate is where an estimate starts, before you add what you see in a specific chart. Trading in probabilities walks through that idea step by step.

Start with the widest reference class that fits. For a stock that opened +20% or more in this period, 65.3% of 671 closed below the open. That is the starting estimate. Then ask whether your conditions match a subgroup with enough observations to matter.

Watch the n before you trust a split. A subgroup of 83 carries far more noise than the full 671. How many trades before this means anything covers how sample size changes what a number can support.

Check the as-of date. These figures stop at 2026-10-02. Markets shift, and the monthly table above already shows a moving premarket share.

Compare against fresh days. The Top Gappers daily recaps list each session's gappers, so you can see how new days line up with these older rates.

Update schedule

We plan to refresh this page quarterly, after each calendar quarter closes. Each update keeps the same definitions, states its own n and as-of date, and labels every figure as a study figure or a FennEdge computation.

Two planned additions have no values yet: how often price touched the prior close or the gap midpoint during the session, and day-two outcomes for every day one, including those that did not gap up again.

Sources

• FennEdge study: What Time Do Small-Cap Gappers Hit Their High of Day? (data as of 2026-10-02)
• FennEdge Top Gappers daily recaps