AI Research SPY

SPY intraday open-to-close returns after large overnight gap-downs (bottom decile)

751
Total trading days

Conventional wisdom says the worst overnight gaps in SPY are bought within the same session — that the opening panic fades into the close. The numbers show why that idea persists. Across 76 bottom-decile gap-down days, the average regular-hours open-to-close return was +0.373%, and 59.2% of those sessions closed in the green.

But the edge is thinner than the average suggests. The median intraday gain was just +0.156%, meaning a few large reversals lift the mean. With a t-stat of 1.772 and a p-value of 0.0805, the bounce doesn't clear the usual 5% significance threshold, and the comparison against the other 675 session days leaves the same impression. It's a lean toward dip-buying, not a proven pattern.

The full analysis below walks through the methodology, the decile table across all overnight-return buckets, and what the wide confidence interval actually allows us to conclude from 76 events.

The research question

For SPY over the past ~3 years, do the largest overnight gap-downs (bottom-decile overnight returns) consistently lead to a positive intraday regular-hours open-to-close return, as dip-buyers rush in to fade the opening panic? Thesis: gap-downs are reliably bought, so the intraday session delivers gains that offset the overnight loss.

How this was measured

Minute SPY bars resampled to daily OHLC. Overnight return = (open / prior close) − 1. Intraday RTH return = (close / open) − 1, representing the regular‑hours move that traders could capture. The bottom decile of overnight returns defines 'gap‑down' days. We then compute the distribution of intraday returns for these gap‑down days, compare it to zero with a one‑sample t‑test, and compare it to all other days with a Welch two‑sample t‑test. A decile table across all overnight‑return buckets reveals the gradient. This is a pure descriptive analysis, not a trading rule.

The key numbers

Total trading days
751
Range: 2023-07-03 to 2026-06-30
Gap-down days (bottom decile)
76
Overnight return ≤ -0.0024
Gap-down mean intraday return
0.3730%
RTH open-to-close
Gap-down median intraday return
0.1562%
Gap-down fraction positive
59.21%
Share of gap-down days that closed higher than open
Other days mean intraday return
0.0163%
N=675
Gap-down vs zero t-stat
1.772
H0: mean(gap-down intraday) = 0
Gap-down vs zero p-value
0.0805
p=0.0805 ≥ 0.05 → no evidence intraday returns are non‑zero on gap‑down days
Gap-down vs other t-stat
1.675
Welch t, positive favours gap-down days
Gap-down vs other p-value
0.0979
p=0.0979 ≥ 0.05 → no statistically‑clear difference from other days

Reading the numbers

On the 76 biggest gap-downs, the average regular-hours gain was +0.37% and 59% closed higher, but the p-value of 0.08 means this could easily be noise — not a clean, reliable dip-buying edge.

The charts

Intraday return: gap-down vs other days
What this chart says

The box plot puts the 76 gap-down days next to the other 675 days. Gap-downs have a higher mean intraday return (+0.37% versus +0.02%), but the gap-down group's observed range runs from -3.29% to +12.02%, so the average is pulled up by a wide positive tail. The two groups overlap heavily, which matches the weak statistical evidence: the difference is not clearly significant (p=0.098).

Overnight return vs intraday RTH return
What this chart says

This scatter plots every trading day's overnight move against its regular-hours return, so the gap-down days are the points on the left side of the chart. If the fade-the-gap thesis were robust, those left-side points would sit mostly above the zero line; instead, the chart shows a wide cloud of points, with one extreme intraday gain of +12.02% towering above the rest. The left tail does not look visibly separated from the rest of the data, which is consistent with the lack of strong statistical significance.

Intraday returns on gap-down days
What this chart says

The histogram of the 76 gap-down days shows most intraday returns clustered around zero, with a long right tail extending to +12.02% and the low end coming in at -3.29%. The mean is +0.37%, but the picture is more a mild tilt than a consistent rebound: only 59% of gap-down days closed above the open. The large positive outliers, rather than a uniform bounce, appear to be doing much of the work.

Intraday return by overnight-return decile

DecileMean intraday returnStdN
P0-P100.00370.018476
P10-P200.00050.010475
P20-P3000.007375
P30-P40-0.00030.008275
P40-P500.00040.007475
P50-P600.00090.006875
P60-P70-0.00060.007375
P70-P800.00050.008375
P80-P90-0.00090.007475
P90-P1000.0010.010475

The takeaway

Not reliably — the gap-down bounce is a lean, not a proven pattern. On the 76 worst overnight gap-down days (bottom decile, overnight return at or below about -0.24%), SPY averaged +0.3730% from regular-hours open to close, and 59.2% of those sessions closed above their open. The median was only +0.1562%, which suggests a few big reversals are doing much of the work. The t-test against zero lands at p=0.0805, and the comparison to the other 675 days (mean +0.0163%) gives p=0.0979 — both above the usual 5% cutoff, so the edge isn't statistically clear. The decile pattern does support the direction: the bottom decile stands out, while the next bucket averages just +0.05% and higher buckets hug zero or go slightly negative. With only 76 events and a bottom-decile standard deviation around 1.84%, the confidence interval on that mean is wide. Practical takeaway: there's a modest dip-buying tendency worth further study, but it's not strong enough to bank on as a standalone rule.

The fine print