Does "Sell in May and Go Away" Work?

We cannot tell you — and that is the honest answer.

The claim as it is usually statedSell in May and go away - Nov-Apr beats May-Oct.

This is not folklore. Researchers found November–April beating May–October in 36 of 37 markets, with UK evidence back to 1694. Testing it properly on our data, 13 of 18 markets lean the documented way — and not one is statistically distinguishable from zero. The reason is arithmetic: year-to-year variation is so large that detecting a 4-point effect in EUR/USD would need about 41 years, and we hold 19. The Dow would need 77 years; we hold 8. A 23-year archive contains 23 observations of an annual effect. It contains 36,005 observations of "EUR/USD in June–August trading hours" — which is why this site measures hours.

What the data shows

Across 19 markets: direction: not testable at this unit — this window comes round once a year, so the sample is 19 occurrences in the median market (5–24 across the nineteen), not the tens of thousands of hourly bars inside them · movement: genuinely quieter in 12 of 19. Median movement in the window, 88% of a normal hour.

Market	In the window	Rest	Difference	Movement	Years
NAS100	52.7%	51.8%	+0.9 pp	96% of normal	7
EUR/USD	49.5%	50.4%	-0.9 pp	90% of normal	24
GBP/JPY	50.5%	51.4%	-0.9 pp	88% of normal	19
ETH/USD	49.4%	50.3%	-0.9 pp	98% of normal	9
UK100 (FTSE)	50.6%	51.3%	-0.7 pp	86% of normal	5
SPX500	52.5%	51.9%	+0.6 pp	83% of normal	7

No significance flag on a season, and here is why. This window comes round once a year. The tens of thousands of hourly bars inside it are not that many independent readings of it — the sample is 19 occurrences in the median market (5–24 across the nineteen). At that size nothing here can clear our bar: detecting the documented "sell in May" effect in EUR/USD would need about 41 years, and we hold 24. The percentages are real and worth reading as description; the column that would tell you they are proven is deliberately absent. What we will not publish →

Six largest differences, whichever way they point. There is no z column on a seasonal claim: it would be a significance flag on a window that recurs once a year, and this site does not print one.

Who says it

Bouman & Jacobsen (SSRN) · Jacobsen & Zhang

How this was measured

Every claim here reduces to a window on the clock or the calendar, so it can simply be checked. We compare the claimed window against the rest of the week on 23 years of hourly data across 19 markets, measuring two separate things the folklore tends to conflate: direction (does the window close higher more often?) via a two-proportion z-test, and movement (is the average move bigger or smaller?). A window can be perfectly ordinary in direction and genuinely quieter in movement — saying which is the entire point.

The bar here is |z| ≥ 2, deliberately more lenient than the |z| ≥ 3 this site applies to its own numbers: when testing someone else's claim, failing to reject it even at a lenient bar is the stronger statement.

Check any hour yourself →

Free on the four major pairs, with the full 23 years and the sample size on every number.