The Full Moon, Revisited: What Survived When I Split It by the Calendar
A while back I ran the full moon against years of Nasdaq data and gave the honest verdict: noise. Full-moon days were, if anything, slightly quieter than normal — nothing like the chaos the folklore sells.
That should've been the end of it. But the full moon drifts — the lunar cycle is ~29.5 days, our months are longer, so the full moon lands on a different date each month. So I asked: does it matter when in the month it falls?
I split every full moon by which calendar week it landed in. At first it looked like I'd found gold.
The pattern that looked too good
Against a baseline of 245.8 points ($492):
| Full moon lands in… | Sample | Median range | vs. normal |
|---|---|---|---|
| Week 1 (days 1–7) | 38 | 231.8 / $464 | −6% |
| Week 2 (days 8–14) | 49 | 285.0 / $570 | +16% |
| Week 3 (days 15–21) | 50 | 242.0 / $484 | −2% |
| Week 4 (days 22–28) | 40 | 188.1 / $376 | −23% |
| Week 5 (days 29–31) | 14 | 195.0 / $390 | −21% |
Look at week 4: a full moon there compressed the range by nearly a quarter, and 75% of those days were pure chop. Week 5 (tiny sample) was even choppier — 93%. The back half of the month, the full moon seems to squeeze the life out of the tape.
I got a little excited. Then I made myself run the control.
The control test (where most people stop, and shouldn't)
The trap: the late month is crowded with other effects — expiration, end-of-month positioning. What if week-4 compression had nothing to do with the moon, and everything to do with late-month days just being quieter? The moon would be a passenger taking credit for the drive.
One way to know: compare full-moon weeks against all days in those same weeks.
| Week | All days | Full-moon days only |
|---|---|---|
| Week 1 | +4% ($513) | −6% ($464) |
| Week 4 | +2% ($499) | −23% ($376) |
This splits the finding clean in two.
Week 4: the moon survived. Week 4 in general is ordinary — +2%, dead average. No late-month compression hiding in the calendar. Yet full moons specifically in week 4 compressed 23%. The calendar doesn't explain that gap. Something tied to the late-month full moon does.
Week 2's +16% is interesting but unconfirmed. It's wide, but I haven't controlled it as tightly, and it could be sample luck. Filed as "watch," not "found."
What I'm actually claiming (and what I'm not)
Narrowly and honestly: a full moon in the last full week of the month is associated with a compressed, choppy Nasdaq session — an effect the calendar alone doesn't account for.
Caveats stay loud: - n=40. Real, not enormous. One odd cluster moves it. I expect it may shrink as data grows — early edges often do. - Correlation, not mechanism. With expiration I can tell you why the range compresses. Here I can't yet. A lead, not a law. - I'm not trading it. Notebook, flagged, watched. Patterns earn a playbook spot by surviving more data.
Why I'm showing you the half that died
Most people would publish the pretty five-row table, call it "The Lunar Range Cycle," and sell a course. I almost believed it myself for an afternoon.
The subtraction is the method. Finding a pattern is easy — the market manufactures coincidences. The work is asking "what else explains this?" and being willing to watch your favorite finding die. Part of this one died. The surviving half is smaller and quieter and honestly held — which makes it worth more than the version that dazzled me.
Sharpen your edge by telling yourself the truth — especially when the truth is "I was half wrong."
— 23HT
How this was done: ~1,530 daily MNQ bars from the 2019 launch through May 2025, tagged by moon phase and calendar week of month, compared by median range against both the all-days baseline and the all-days-in-that-week control.