How often do opening gaps in US stocks fill?

01Question

"Gaps get filled" is one of the most repeated lines in trading. It is usually stated as a tendency: a stock that opens above yesterday's close will, more often than not, trade back down to it. We ask two narrower questions. How often, and how soon, did US stocks reach the previous close after opening away from it? And did they reach it more often than they reached any other level the same distance away?

The second question is the one that matters. A small gap is a short distance, and stocks travel short distances every day. If a gap fills no more often than price wanders the same distance in the other direction, the previous close is not doing anything special.

02What we measured

We took every regular session (09:30 to 16:00 New York) for the 154 stocks and 53 ETFs in the SixOne8 US equity catalogue, from 1-minute candles. A session has a gap when its 09:30 open differs from the previous session's 16:00 close. We size each gap in multiples of the stock's 14-day average true range (ATR) as it stood at that close, so a gap on a quiet utility and on a volatile chip maker are measured against what each normally moves. Sessions whose gap is under 0.05 ATR count as no gap.

A gap fills when any 1-minute bar in the session reaches the previous close. For each gapped session we also record the mirror level, the same distance from the open on the other side, and whether price reached it. And as a second baseline, we measure how often sessions without a gap travelled the same distance from their open in either direction.

03How often gaps filled

Of 168,014 gapped sessions, 58% reached the previous close the same day. The median gap was 0.26 ATR, about 0.58% of the price. Figure 1 shows the rate falling smoothly with size: 87% for the smallest band, 64% at 0.2–0.3 ATR, 30% at 0.5–0.75 and 7% above 1.5. There is no size at which gaps stop filling; the likelihood simply thins out.

Figure 2 lays the same sessions out as a hundred squares. For gaps under half an ATR, 68% reached the previous close, and 21% of those fills came inside the first minute. For larger gaps, 77% never got there.

The blue line in Figure 1 barely separates from the grey one. For every band up to half an ATR, the mirror level was reached about as often as the previous close. Table 1 gives the difference and its interval for each band.

04Is the previous close special?

Pooling all gaps under half an ATR (133,320 sessions), the previous close was reached 67.7% of the time and the mirror level 66.8%: a difference of +0.8 pts, with a 90% interval of −0.3 to +2.0 points. For these gaps, which are four in five of all gaps, the previous close behaved like any other level at that distance.

Larger gaps are different, modestly. Above half an ATR (34,694 sessions) the previous close was reached 23.0% of the time against 20.2% for the mirror, +2.8 pts (+0.7 to +5.0). Large gaps partly retrace more often than they extend by the same amount. The effect is a few points on a base of one in four. Figure 3 shows it in nearly every year for these larger gaps, while for smaller gaps a difference appeared in the first year and had faded or reversed by the last.

The no-gap baseline (the teal line in Figure 1) sits close to both for small gaps and below them for large ones. That comparison flatters the gap for a reason unrelated to the previous close: days that open with a large gap are more volatile days, so price travels further on them in every direction. The mirror level shares the session's volatility, which makes it the fairer baseline.

05When fills happened

Small gaps filled fast. Of the gaps under 0.1 ATR, 39% reached the previous close inside the 09:30 minute itself, and across all gaps 19% of fills came in that first minute. A rule that decides at the close of the first bar and fills at 09:31 never sees those (Figure 4). Figure 5 follows the rest through the day: the median fill came 14 minutes after the open for gaps of 0.2–0.3 ATR and 74 minutes for 0.5–0.75.

Each line flattens as the day goes on. A gap still open at midday had a small chance of filling by the close, whatever its size.

06Sizing a gap: ATRs or percent

A 1% gap is routine for a volatile stock and rare for an index fund. Measured in percent, the same gap size means very different things on different instruments, so fill rates spread widely between them. Measured in ATRs, they cluster. Figure 6 shows the fill rate for gaps of 0.2–0.5 ATR on each instrument: most sit within a few points of the median, 56%.

Table 2 puts a number on it. Across four bands of matched size, the spread of fill rates between instruments averaged 6.6 points with ATR sizing and 14.9 points with percentage sizing. If you write a gap condition, size it in ATRs.

07Stocks, ETFs and direction

ETFs filled less often than single stocks at the same ATR distance, and the gap widened with size: 26% against 33% at 0.5–0.75 ATR. Gaps up and gaps down filled at close to the same rate in every band. About half of the sessions that filled went on to close beyond the previous close, giving the whole gap back by 16:00.

08The path before a fill

For sessions that filled after the first minute, the median move away from the previous close before the fill was about 0.12 ATR, whatever the size of the gap. The tail grows with size: one in ten of the 1.0–1.5 ATR gaps that filled first moved 0.58 ATR the other way (Figure 7). This describes the path; it does not make a gap a trade. A rule that waits for a fill takes the moves against it too, and on most large gaps the fill never comes.

09Methodology

Every figure in this paper comes from the definitions below. Where we made a choice, we state it.

Data

1-minute candles from Twelve Data, stored by SixOne8, regular sessions only (09:30–16:00 New York). Prices are split-adjusted, not dividend-adjusted.

Universe

All 207 US stocks and ETFs in the SixOne8 catalogue: 154 stocks and 53 ETFs. No instrument was removed after the fact; instruments listed during the window enter when they have 14 sessions.

Window

Sessions from 15 November 2022, the first with a complete 14-day ATR, to 30 September 2026: 196,278 sessions on 971 trading days.

Gap

09:30 open minus the previous session's last close, divided by the 14-day ATR (Wilder's smoothing, the same ATR SixOne8 computes) at that close. Under 0.05 ATR counts as no gap.

Fill

The first 1-minute bar whose high (gap down) or low (gap up) reaches the previous close.

Mirror

The level at the same distance from the open in the direction of the gap.

No-gap baseline

For sessions without a gap, the share that travelled at least a given number of ATRs from the open, either way, evaluated at each gapped session's own gap size and averaged.

Intervals

90% intervals from 1,000 bootstrap resamples of whole trading days, since gaps across stocks share the same news days.

Excluded

1,220 sessions with fewer than 180 minutes (early closes and data gaps), 1,476 with no 09:30 minute, 142 after a break in the stored history, and 1 with an implausible price.

10Limitations

Four years is a short window for stocks and covers one broad market regime. We do not separate earnings or news gaps from the rest, and many of the largest gaps are earnings. Dividend days open lower by the dividend and are counted as small gaps down. A fill is measured from 1-minute highs and lows, so a level touched for a moment counts the same as one traded through.

Above all, a fill rate is not a profit. The share of gaps that filled says nothing about what a rule fading them would have earned after the moves against it, the gaps that never filled, and costs. That is a separate test, and it is the one to run before relying on any of this.

11Test it yourself