Do Price Gaps Get Filled? We Tested 1,748 NQ Opening Gaps — the Fill Rate Is a Distance Illusion

A price gap is the simplest object in trading: today’s open minus yesterday’s close. And from that one number, three simple questions follow — the kind of reasoning that should precede any strategy:

1. Is direction predictable after the gap?
2. Are gaps predictable from prior price action?
3. Do gaps get filled?

We answered all three on 1,748 NQ sessions (2019–2026), with the control group almost no gap study bothers to run — and then tested the strategy people actually trade: look at the gap, wait, and see whether the market continues or starts working the gap off.

One stat does the judging below: t (the t-statistic) — how far an average sits from zero relative to its own noise. |t| > 2 is a real effect; less is indistinguishable from luck.

Q3 first, because it’s the famous one: do gaps fill?

Gap size (vs 14-day avg range)	Filled same session	Mirror level touched (same distance, no story)
< 5%	96%	96%
5–10%	87%	89%
10–20%	78%	77%
20–40%	60%	60%
> 40%	33%	26%

The mirror control is the whole story. For every gap we also checked a placebo level the same distance from the open in the opposite direction — a price with no story attached. The touch rates are identical for four of five buckets. “93% of gaps get filled” — the number the courses quote — is true and empty: price touches any level that close to the open, because the market’s daily wander is bigger than a small gap. Distance, not memory.

The honest nuance: the largest gaps (>40% of a day’s range) do show a genuine pull — 33% filled vs 26% for the placebo, about +7 points of true gap-directed attraction (z ≈ 2.8). The market does lean back toward big gaps slightly more than chance. Hold that thought until the P&L section.

And the folklore’s favorite escape hatch — “every gap fills eventually” — is just false: of the gaps that didn’t fill same-day, 46% were still unfilled five sessions later.

Q1: is direction predictable after the gap?

No. Correlation between the gap and the rest of the day: −0.06. No gap-size quintile reaches significance in either direction (the biggest up-gaps lean toward fading at t = −1.5 — suggestive, not real). Traded naively from the open, both classics fail:

• Gap-and-go (follow the gap, hold to close): −$104k per contract over 7 years (t = −0.9).
• Fade to the fill (target = prior close, stop beyond the open): −$7k (t = −0.1) — a coin flip paying commissions. Its only good stretch is 2024–26; the 2019–23 train period is negative. We killed the same trade once before in our knowledge-base sweep for exactly this reason: a holdout-only “edge” is a regime, not a strategy.

Q2: are gaps predictable from prior action?

Size yes, sign no. Tomorrow’s gap magnitude correlates +0.32 with today’s range — volatility clusters, as it does everywhere (the one survivor of our day-type research). But the direction contains nothing: P(gap up | up day) = 55.7% against a 56.0% base rate. You can know a big gap is coming; you cannot know which way. A forecast of magnitude without direction sizes positions — it doesn’t create them.

The strategy everyone actually trades: wait and see

Nobody trades the open blind — the practitioner version is: watch the first minutes; if the market keeps pushing in the gap direction, go with it; if it starts working the gap off, trade the fill. We tested exactly that, K = 15/30/60 minutes:

• The tilt is real: if the first hour moves with the gap, the day closes with the gap 60.2% of the time (vs 52% unconditional). If it moves against, the day closes against 55.6%. Waiting genuinely improves the odds.
• The money isn’t: continuation-after-confirmation earns +$113k at t = +1.8 — under the significance bar, and the holdout alone is t = +0.7. The fill-trade side is noise at every K (at K = 60 it loses, t = −1.8).
• The control is the punchline: the identical wait-for-the-first-hour rule on days with no gap (<5% of range) made more money (+$118k, t = +2.3) on a fifth of the trades. The information is in early-session momentum itself. The gap contributes nothing — if anything it dilutes it.

The bottom line

Three simple questions, honestly answered: direction after a gap is not predictable (−0.06); gaps are predictable in size but not sign; and the famous fill rate is a distance illusion — a placebo level with no story gets touched exactly as often, except for the very biggest gaps, whose small genuine pull still doesn’t survive costs. The wait-and-see refinement improves probabilities and not P&L, and its own control group outperforms it without the gap.

That’s not a disappointing result — it’s the method working. The same three-question discipline that kills the gap trade is the one that occasionally says yes: the last submission that survived it is running live on our board right now. Most simple concepts die under a control group. The point of asking simple questions is finding the few that don’t.

Methodology: NQ continuous front-month, 1-minute RTH bars 2019–2026 built from real trade prints, 1,748 sessions. Gap = 09:30 ET open minus prior 16:00 ET close, normalized by the trailing 14-day average daily range. Fill = touch of the prior close within the same RTH session; mirror control = touch of open + gap-distance in the opposite direction. Direction tests on open→close returns, daily-aggregated t-stats, 2019–23 train / 2024–26 holdout. Trades costed at $4.50 commission + 2-tick slippage per round trip. Wait-and-see: first-K-minute move from the open as trigger, continuation held to close, fill-trade targeted at the prior close with a half-gap stop.