marketcorollary

Study two · the folklore battery

The Almanac of Everything Else

Every way the world is said to move the market — the moon, the planets, eclipses and comets, groundhogs and cicadas, popes and quarterbacks, skyscrapers and shutdowns — re-tested from raw prices, 1928 to the present. Event calendars are rebuilt from first principles (astronomical ephemeris, the Umm al-Qura calendar, the tz database, the GFZ geomagnetic record). Each daily indicator faces a 2,000-draw circular-shift permutation test; the whole battery then passes through Benjamini–Hochberg false-discovery correction. make it weird was the brief; almost everything is noise is the finding.

The sky over Wall Street tonight

Computed from the ephemeris on every rebuild. Decorative and, as far as we can measure, economically inert.

Forthcoming omens

None of this predicts returns. We checked. It's the whole point of the page.
Signals on trial
35
27 daily + 8 annual rules
Survive FDR correction
1
Santa / turn-of-year — untradeable
Super Bowl accuracy
53%
vs 67% by always guessing up
Data behind the verdicts
98y
S&P daily · 141y Dow · live-refreshed

The Sky ninety-eight years of markets vs. the heavens


ⅩⅢThe Calendar days of dread, windows of cheer


Beasts, Rites & Omens oracles with fur, shells and smoke


The Works of Man stadiums, shutdowns, skyscrapers


The Full Ledger every number behind every verdict


The daily battery

Diff = event-day minus control-day mean return (bps/day); Ann annualizes it. p(perm) is the permutation p-value; q(FDR) is false-discovery-corrected. Click a header to sort.

A SURVIVES verdict means q<0.05 after correction; suggestive means raw permutation p<0.05 but not after correction; noise is everything else.

Annual rules & event rules

Tested against the market's base rate — it rose in 67% of years, so a rule must beat always guessing up, not a coin flip.

🝳The Untestables lovingly catalogued, honestly abandoned


Some famous indicators cannot be backtested honestly — the data never existed, died with a trade association, or was a joke from birth. They are preserved here as specimens.

Hemline Index 1926

Skirts up, stocks up. A century of retellings; no dataset has ever contained both hemlines and dates.

Lipstick Index 2001

Recessions sell lipstick — except 2008, when lipstick fell 7%, and 2020, when the index quietly became skincare.

Hot Waitress Index 2009

New York Magazine's contribution to macroeconomics. No register of waitress hotness exists, mercifully.

Necktie Indicator lore

Tie sales were said to rise with job anxiety. The US tie-makers' association disbanded in June 2008 — months before Lehman. The indicator died before the recession it existed to predict.

Boston Snow Indicator lore

A white Christmas in Boston means a green year on Wall Street. Traders note its initials.

SI Swimsuit Indicator 1978

American cover model, American gains — ~89% hit rate at peak fame. 2008: American model. S&P −38%.

The McRib Effect 2016

S&P +7 bps on days the McRib is available (p=0.19). Built and published as a parody of exactly this page.

Butter in Bangladesh 1995

R² = 0.99 against the S&P with cheese and sheep added. The undefeated world champion of spurious correlation, published as satire.

Sotheby's Stock n≈4

Auction manias marked the tops of 1989, 1999 and 2007. The series went private in 2019, taking the indicator with it.

The fault, dear Brutus, is not in our stars —
it is in our sample size.

Questions this answers

Is there a Santa Claus rally?
It is the one folk pattern that survives correction: the turn-of-year window averages +1.60% over ~7 trading days and was positive in 77% of 98 years — but it is calendar-internal seasonality, not a tradeable external signal, and it has missed several recent years.
Does Friday the 13th cause stock market crashes?
No. Friday the 13th is statistically indistinguishable from other Fridays, and shows the classic data-mined signature: negative before its famous publication, positive after. Even Friday the 13th during Mercury retrograde — the most cursed day astrology can assemble — has averaged positive returns.
Does Mercury retrograde affect the stock market?
No. Across nearly a century of daily data the retrograde-vs-direct difference is a fraction of a basis point. Venus and Mars retrogrades — our control planets — are equally empty.
Does the World Cup hurt the stock market?
The tournament window is negative on average against the same dates in non-Cup years, replicating the published 'loser effect' direction — but it does not survive false-discovery correction.
Is the skyscraper curse real?
Suggestive but tiny: after the eight world's-tallest-building completions in our sample, the -15.0% average return over the following year compares to +8.2% (median +9.5%) unconditionally — but n=8 and p≈0.29. The Chrysler Building's secret spire did go up one day before the 1929 crash.
Do geomagnetic storms affect stock returns?
An Atlanta Fed paper (2003) said yes — lower returns in the six days after storms. Our replication on GFZ Ap data through the present finds the same direction, but it is not statistically distinguishable from noise after correcting for the size of this battery.