“Gaps always fill” is the most repeated claim in retail trading and one of the least examined. It survives because it is almost true in a useless way — given enough time, most price levels get revisited. The version traders actually act on is narrower and far shakier: that a gap will fill today, and that the odds are good enough to trade against.
Put real numbers on it and the claim splits apart along two seams. The first is gap size, where gap fill probability runs from roughly 78% down to 8% inside the same dataset. The second is time — the headline rate itself drifts enough between rolling windows that quoting a single number is close to meaningless. Here is what the published data shows, and why an overnight trader should read it differently than a day trader does.
What the Statistic Actually Measures
A gap is the difference between today’s regular-session open and yesterday’s regular-session close. A fill is price returning to that prior close at some point before the bell. Most published studies measure exactly one thing: whether a 100% fill happened by 4:00 p.m. That is a day trader’s question. It quietly assumes you are at the screen from the open, and it says nothing about the overnight risk you already took to get there.
Worth separating out too: the gap forms against the prior cash close, not against wherever futures were trading at 3 a.m. That distinction matters more than most traders expect, and we unpacked it separately in why the overnight futures print is not your gap.
Gap Fill Probability by Size: 77.8% Down to 8.2%
The largest published breakdown available covers 2,791 NQ futures sessions from January 2015 through December 2025, compiled by TradingStats. The blended headline: 60.3% of gaps fill completely by the close. Sorted by size relative to the 14-day ATR, that average dissolves.
| Gap size | Sessions | 100% fill by close |
|---|---|---|
| Tiny (< 0.3x ATR) | 1,711 | 77.8% |
| Small (0.3–0.7x ATR) | 690 | 42.0% |
| Medium (0.7–1.2x ATR) | 241 | 25.6% |
| Large (> 1.2x ATR) | 149 | 8.2% |
Nearly a tenfold spread. And notice where the sample sits: 61% of all sessions fall in the “tiny” bucket, which is why the blended 60.3% looks reassuring. The average is mostly measuring days when the market barely gapped at all.
The perverse part for anyone holding a position overnight is that the gaps you care about are precisely the ones that don’t fill. Nobody loses sleep over a 0.1% gap. The 1.5% gap against you is the one you want reversed — and that is the bucket sitting at 8.2%. Independent equity data points the same direction: SharePlanner’s analysis of SPY and QQQ found roughly 45% of 1%–1.99% overnight gaps filled same-day, dropping to about 30–33% once the gap exceeded 2%.
The Headline Rate Is a Moving Target
Here is the part almost nobody publishes, because it undercuts the premise of publishing a number at all. Trade That Swing has updated the same rolling six-month SPY gap study three times since August 2025, using Edgeful’s data. The number keeps sliding.
| Snapshot (trailing 6 months) | Gap-up fill | Gap-down fill |
|---|---|---|
| August 2025 | ~60% combined | |
| March 2026 | 59% | 69% |
| July 2026 | 53% | 59% |
Ten points of drift in under a year, on the same instrument, from the same data provider. That is not a measurement error — it is the market changing. The 2025 window sat inside a steady uptrend where small gaps down got bought reflexively. As the regime shifted, so did the fill rate. Any “gap fill probability” you saw quoted in a 2025 article has already expired.
The Weekday Splits Are Mostly Noise
If the headline moves, the sliced-thinner numbers move violently. Across those same three snapshots, the SPY Monday gap-up fill rate read 54%, then 20%, then 45%. Friday gap-downs read 20%, then 78%, then 30%. Those swings are sample size, not signal — a six-month window contains roughly 26 Mondays, and splitting them by gap direction leaves a dozen observations per cell.
The long-horizon data is calmer and duller, which is the point. Over 2,791 NQ sessions the weekday spread compresses to Monday 53.9% at the low and Wednesday 63.5% at the high. Real, small, and nothing to build a rule around.
A Fill Is an Opening-Hour Event
The timing data reframes the whole statistic. Of NQ gaps that filled, 43.3% completed inside the first hour. The rest of the session — five and a half hours — contributed 16.9% between them, and 39.7% never filled at all.
| Fill level | By 10:30 a.m. | By noon | By close |
|---|---|---|---|
| 25% of gap | 72.9% | 80.2% | 86.6% |
| 50% of gap | 55.5% | 65.5% | 76.1% |
| 100% of gap | 43.3% | 51.9% | 60.3% |
Median time to a complete fill is 18 minutes. For tiny gaps it is 7 minutes. Whatever mean reversion exists is spent almost entirely before most overnight traders have finished their first coffee — and the afternoon adds only 8.4 percentage points.
The gap fill is not a safety net you hold a position into. It is a fast, front-loaded event that mostly resolves before 10:30 a.m. and mostly happens to gaps too small to have hurt you.
Reading This as an Overnight Trader
Four things survive the disaggregation and are worth carrying into a position:
- Size dominates direction. Gap-downs fill 62.2% versus 58.8% for gap-ups — a real but small 3.4-point edge. The size gradient spans 70 points. Sort by size first, always.
- Never size a position assuming the gap comes back. An 8.2% fill rate on the large-gap bucket means the recovery you are counting on arrives roughly one morning in twelve. Size to the loss you can absorb at the open instead — the arithmetic is in the overnight math most traders get backwards.
- Volatility regime tilts the odds against you exactly when it matters. Fill rates run 63.2% in low-volatility conditions and 56.3% in high-volatility ones. Higher volatility also produces bigger gaps, so the two effects compound.
- The drawdown before a fill is real. Median adverse excursion on a gap-fade trade is 0.34x ATR, but the 90th percentile reaches 0.97x — a full day’s range against you before the fill, if it comes. One in ten sessions whipsaws.
None of this makes gap statistics useless. It makes them context, not a plan. The same caution applies to our own pressure readings: they are decision support for structuring a position before the close, not a prediction that the morning will cooperate. And the deeper question of whether overnight exposure is worth holding at all is one we looked at separately in what 8,450 nights of SPY data actually show.
The honest summary of gap fill probability is this: a well-defined number for small gaps, a coin flip for medium ones, a near-certain no for large ones, and a figure that drifts by ten points a year regardless. Anyone quoting you one percentage is averaging away the only part that matters.
This article is for educational and informational purposes only. It does not constitute financial advice, investment recommendations, or a solicitation to buy or sell any securities. Past performance and historical data do not guarantee future results. All trading involves risk, including the potential loss of principal. Always do your own research and consult with a qualified financial advisor before making trading decisions.
