AI Research COIN

COIN Monday-morning gaps vs ordinary overnight gaps (last ~3 years)

Monday opens for Coinbase have been noticeably larger in size than ordinary overnight moves, but they don’t carry a reliable directional signal. Over the last ~36 months the mean absolute Fri→Mon gap was 2.433% versus 1.800% for Tue–Fri gaps (Welch p = 0.0034), while the average signed Monday move (+0.248%) is small and statistically indistinguishable from zero. In short: weekend crypto repricing piles up into bigger openings, not into a predictable buy-or-sell bias.

This study compares COIN’s regular-session open to the prior regular close for each trading day, partitioning Fri→Mon gaps against Tue–Fri overnights and testing magnitudes (Welch t-test) and directionality (one-sample tests, sign fractions). The detailed statistics, charts, and robustness checks that support these conclusions follow below.

The research question

For COIN over the past ~3 years, are Monday-morning gaps (Friday close → Monday open) systematically larger in magnitude than ordinary weekday overnight gaps — COIN repricing to a weekend of nonstop crypto trading it couldn't participate in — even though the gap's direction carries no reliable edge? Thesis: Monday gaps run meaningfully wider than other overnight gaps because two-plus days of crypto moves pile up while the stock is shut, but they come out roughly symmetric, making it a volatility quirk rather than a tradeable directional signal.

How this was measured

Filtered COIN minute bars to US regular trading hours (Mon–Fri, 09:30–16:00 New York). For each trading day, took the RTH open (first 09:30 bar) and prior day’s RTH close (last 16:00 bar) to compute the overnight gap: gap = open_t / close_{t-1} − 1. Partitioned gaps into Monday (Fri→Mon) vs ordinary weekdays (Tue–Fri). Compared absolute-gap distributions with a Welch two-sample t-test; assessed directional bias via one-sample t-tests of signed gaps against zero and the fraction of positive gaps. All computations are within the last ~36 months of available data.

The key numbers

Monday mean |gap|
2.4334%
N=142 Mondays
Tue–Fri mean |gap|
1.8000%
N=608 weekdays
Welch t (|gap| Mon vs Tue–Fri)
2.972
Positive favors Monday
Welch p-value (|gap|)
0.0034
p=0.0034 < 0.05 → Monday gaps are wider
Monday mean signed gap
0.2482%
N=142 Mondays
One-sample p (Mon mean ≠ 0)
0.3873
p=0.3873 ≥ 0.05 → no clear directional bias
Monday fraction positive
54.225%
Share of Mondays with open > prior close
Tue–Fri mean signed gap
0.2132%
N=608 weekdays
One-sample p (Tue–Fri mean ≠ 0)
0.0363
Welch p (Mon vs Tue–Fri signed means)
0.9082

Reading the numbers

Monday absolute gaps average 0.024333891328212374 (N=142) versus 0.01800025516896964 for Tue–Fri (N=608); the Welch test p=0.0033703133430484575 says Monday gaps are significantly wider in magnitude. The mean signed Monday gap is 0.00248224395338233 with p=0.3872822113135215, so there is no reliable directional edge for Mondays.

The charts

COIN absolute overnight gaps: Monday vs Tue–Fri
What this chart says

This box plot puts the size of Friday→Monday absolute gaps next to ordinary overnight gaps and the whole Monday box is shifted upward: the Monday mean absolute gap is 0.0243 (n=142) versus 0.0180 for Tue–Fri (n=608). Look at the long upper whisker and the much larger Monday max (0.2072) compared with Tue–Fri max (0.1413) — Mondays show a longer tail of big moves. That visual plus the sample counts lines up with the statistical result that Monday gaps are meaningfully wider in magnitude.

COIN signed Monday gaps (Fri→Mon)
What this chart says

The histogram of signed Friday→Monday gaps spans both directions: the min is −0.2072 and the max is 0.1092 with mean 0.00248224395338233 (n=142). Note the extreme negative outliers (down to −0.2072) while the bulk of bars sit near zero; despite a slight majority of positives (fraction positive 0.5422535211267606), the one-sample p for the Monday mean is 0.3872822113135215, so the average signed gap is not reliably different from zero. In short, Monday gaps are often larger in magnitude but not consistently biased up or down.

COIN signed Tue–Fri overnight gaps
What this chart says

The Tue–Fri signed-gap histogram covers a similar central band with min −0.1125, max 0.1413 and mean 0.0021316751855639296 (n=608). The one-sample p for the Tue–Fri mean is 0.03628900153707452, indicating a small positive average across ordinary weekdays, but the Welch p comparing Monday vs Tue–Fri signed means is 0.9082371989224218, meaning Monday and other weekdays show no meaningful difference in direction. Combine that with the significant difference in absolute gaps (p=0.0033703133430484575) and the pattern is clear: Mondays are wider in magnitude but not directionally distinct.

Overnight gap summary: Monday vs ordinary weekdays

groupNmean_abs_gapmedian_abs_gapstd_abs_gapmean_signed_gapmedian_signed_gapfraction_positive
Monday (Fri→Mon)1420.02430.01860.02390.00250.00440.5423
Tue–Fri (prev→open)6080.0180.01370.01750.00210.00050.5132

The takeaway

Yes — Monday (Fri→Mon) opens are meaningfully bigger in size but not a reliable directional trade. The mean absolute Monday gap is 2.433% vs 1.800% for Tue–Fri, with N=142 Mondays and N=608 ordinary overnights and a Welch p=0.0034 (about a 3-in-1,000 chance this size difference is random). Signed Monday gaps average +0.248% (p=0.387, i.e., roughly a 39-in-100 chance this is noise) and 54.2% of Mondays are up, so there’s no clear directional edge from the weekend. Ordinary weekdays show a tiny signed mean (+0.213%) but the Mon vs Tue–Fri signed-mean comparison is p=0.908, which strongly supports no directional difference between groups. Bottom line: the weekend piles up crypto moves into larger openings on Mondays — a volatility/size quirk you can price or hedge around, but not a dependable long/short signal.

The fine print