Most published SPY gap statistics arrive with a single number attached — 60%, 70%, 76%, depending on who is talking — and no date. “Gaps always fill,” the claim goes. Nobody ever tells you when that number was measured, and that turns out to matter more than the number itself.
Cory Mitchell, CMT, has been publishing a rolling six-month study of SPY gap fills on Trade That Swing, using data from Edgeful, and — usefully for our purposes — he has left the old snapshots in place instead of overwriting them. That gives us three readings of the same methodology, on the same instrument, roughly a year apart: August 2025, March 2026, and July 2026.
Reading them side by side is the closest thing to a stability test that retail traders have easy access to. Some of these statistics barely moved. Others moved so much they should probably be thrown out.
The headline number drifted about 10 points in a year
Start with the top-line fill rate — the number most people quote.
| Snapshot (trailing 6 months) | Gap-up filled | Gap-down filled |
|---|---|---|
| August 2025 | ~60% combined | |
| March 2026 | 59% | 69% |
| July 2026 | 53% | 59% |
Gap-down fills went from 69% to 59% in under five months. That is not a rounding error — it is the difference between a setup worth taking and a coin flip with commissions attached. If you built a plan in March on “roughly seven in ten,” by August you were running it into a materially different distribution.
Worth noting what a “fill” even means here: price gaps away from the prior close and then touches that prior close at some point during the regular session. Change the definition to a 75% fill or a 120% fill and every number on this page moves. That alone explains a good chunk of why published SPY gap statistics disagree so wildly with each other.
What held up: fill rate by gap size
The size relationship is the sturdiest thing in the dataset. Small gaps fill often, large gaps fill much less often, and that ordering never broke across any of the three snapshots.
| Gap size | Aug 2025 | Mar 2026 | Jul 2026 | Spread |
|---|---|---|---|---|
| Gap down, 0–0.19% | 92% | 88% | 83% | 9 pts |
| Gap up, 0–0.19% | 87% | 79% | 82% | 8 pts |
| Gap down, 0.2–0.39% | 69% | 79% | 65% | 14 pts |
| Gap up, 0.2–0.39% | 53% | 62% | 58% | 9 pts |
| Either, above 0.4% | below 50% | 29–36% | 37–50% | — |
The tightest bucket in the whole study is the smallest one: a sub-0.2% gap down filled 83–92% of the time in every single snapshot. That is a real, persistent structural feature — a gap that small sits inside normal opening noise, so it barely requires a directional move to close.
The practical consequence runs the wrong way from how most people trade it, though. The gaps with the highest fill probability are the ones offering the least profit, and the gaps big enough to be worth trading are exactly the ones that fill less than half the time. If you are carrying a position overnight, that asymmetry is the whole story — and it is why SPY and QQQ do not carry the same overnight gap risk despite tracking each other closely intraday.
What fell apart: the day-of-week splits
Now the same study, sliced by weekday. This is the cut you see quoted constantly on trading Twitter — “never fade a Monday gap,” that sort of thing.
| Gap-down fill rate | Aug 2025 | Mar 2026 | Jul 2026 | Swing |
|---|---|---|---|---|
| Monday | 70% | 78% | 85% | 15 pts |
| Tuesday | 55% | 79% | 54% | 25 pts |
| Wednesday | 80% | 80% | 62% | 18 pts |
| Thursday | 60% | 25% | 57% | 35 pts |
| Friday | 20% | 78% | 30% | 58 pts |
Friday gap-downs filled 20% of the time, then 78%, then 30%. Thursday went 60%, 25%, 57%. The gap-up table is no better — Monday reads 54%, then 20%, then 45%.
A statistic that swings 58 points between measurements is not describing a market behavior. It is describing a small sample.
Why the weekday numbers move so much
Run the arithmetic. Six months is roughly 126 trading sessions. Split those across five weekdays and you have about 25 of each. Split again by gap direction and each cell in that table rests on something like a dozen observations.
At twelve observations, one session flipping from fill to no-fill moves the percentage roughly eight points. Three sessions going the other way moves it twenty-five. The Friday column is not telling you Fridays changed — it is telling you that a handful of Fridays landed differently, which is exactly what you should expect from a sample that size.
The more finely you slice a dataset, the more precise the output looks and the less it actually means. Precision and reliability are not the same property.
This is the same discipline we try to apply to our own numbers. When we audited 24 pressure readings and reported 14 correct, the honest caveat was that 24 observations is a small sample too — good enough to describe a stretch, nowhere near good enough to promise a hit rate.
The most durable number in the study is about timing, not direction
One figure held across the full year better than anything else. In the August 2025 snapshot, more than 80% of gaps that filled did so by noon Eastern. In July 2026, that figure was 89%.
That is a structural statement about when liquidity and information get absorbed, and it survives regime changes far better than any directional percentage. The overnight session builds an imbalance; the first two and a half hours of the cash session resolve most of what is going to resolve. If you want the mechanics of how that imbalance forms in the first place, we walked through what actually sets SPY’s opening price while you’re asleep.
Practically: if a gap has not filled by lunch, the odds it fills later that session drop sharply. That is a cleaner decision rule than anything in the weekday table.
How to use SPY gap statistics without fooling yourself
- Ask when it was measured. A gap-fill rate with no date attached is unusable. This one moved ten points in five months.
- Ask how “fill” was defined. 100%, 75% and 120% fills produce completely different tables from identical price data.
- Count the cells. If a statistic is sliced by weekday, direction and size at once, estimate the observations behind it. Under ~30, treat it as a story rather than an edge.
- Prefer stable structure over precise trivia. “Small gaps fill more often than large ones” and “most fills happen before noon” both survived a year. “Fade Friday gap-downs” did not.
- Note the regime. All three snapshots came from an S&P 500 uptrend, which is part of why gap-downs filled more readily than gap-ups. In a downtrend that asymmetry may well invert.
None of this makes SPY gap statistics useless. It makes them context, not instructions. That is roughly how we intend SpyTomorrow’s nightly pressure reading to be used as well — as one structured input you commit to before the open, sized appropriately, rather than a number that tells you what tomorrow does. The traders who get hurt by statistics like these are rarely the ones who ignored the data. They are the ones who took a twelve-observation cell and sized it like a certainty.
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.
