What 259 market theses taught us
We wrote down 259 beliefs that traders act on and tested every one of them on EquationDB's stored history, with the same rule for each: the edge over an ordinary day, at the horizon the belief names, with t of 2 or more to count as supported. 74 were supported, 21 were mixed and 164 were not. Here is what stood out.
Breakouts lag
Large caps breaking to a new 52-week high trailed their own ordinary 20-day return by 0.31% over 12,421 occurrences since 2000 (t ≈ −4), and the gap was negative in every decade. Golden crosses, MACD crosses, Supertrend flips, Ichimoku breakouts and three white soldiers did no better: most showed a small negative edge over 20 to 60 days in large caps.
Fear has paid
When the VIX closed above 30, SPY, QQQ, IWM and DIA returned 3.41% more than usual over the next 60 days (t ≈ 6), and the edge held in all four decades tested. The mirror belief, that the VIX falling back below 20 is the all-clear, came out the other way: −1.68% (t ≈ −4.6).
Low volatility did not win in large caps
Among large caps, low-volatility months returned 0.60% less than usual (t ≈ −8). Stocks with realised volatility above 60% returned 2.56% more (t ≈ 5.6). Defensive sector funds (utilities, staples, health care, low-volatility and dividend funds) did not beat SPY over the 20 days after a 2% down day. None of the gaps was significant, but none favoured the defensive side.
Earnings: misses drift, beats no longer do
After a miss, top-500 stocks kept drifting down: −1.42% over 60 days (t ≈ −6.3), negative in 19 of 22 years. After a beat the drift is gone: +0.04%. Beats of 20% or more still drifted more than small beats (t ≈ 3.9).
The calendar: some effects are real
The first trading day of the month (+0.18%, t ≈ 4.4) and Turnaround Tuesday (+0.50%, t ≈ 6.6) passed. September was weak (−1.78%, t ≈ −2.9). Sell in May and the Halloween indicator pointed the right way without being reliable. Mondays were not weak.
Big gap-downs did not keep sliding
Large caps that gapped down 10% or more were up 7.09% more than usual 60 days later. This is the result most flattered by survivorship: the universe is today's top 500, so the companies whose gap-down was the start of the end are not in it.
Weather moved less than the stories say
None of the cold-snap tests on natural gas (Chicago, New York, Boston, Houston, Minneapolis) showed a reliable edge. After Boston blizzards, retailers outperformed, the opposite of the belief.
What we take from it
Most popular rules are not supported once they are measured against an ordinary day rather than against zero. A few survive, and they tend to be about fear and about bad news, not about strength. Every one of the 259 is on Explore › Thesis with the statement behind it, so you can test the result rather than take our word for it, and write your own.
Method: OUTCOMES on stored daily history; edge = mean forward return after the signal minus the same symbols' unconditional return over the same years; t = edge ÷ (standard deviation ÷ √n); universes are today's members (survivorship); results as of 2026-10-16.