Testing popular trading rules
We take trading rules people repeat on forums and score them against DawnScan's nightly scan logs of ~6,000 US stocks, delisted stocks included. Each time a rule fired, we check whether the stock was up 20 trading days later, and compare that with stocks of similar volatility on the same day, not with the whole market. Otherwise volatile stocks look like winners just because they move more.
Rules and pass criteria were fixed before looking at results. A Benjamini–Hochberg correction is applied across all rules, and results are recalculated every Sunday. "Edge confirmed" requires both a higher up-rate and a mean excess return whose confidence interval stays above zero.
Summary · 24 rules · 377,194 cases
- Edge confirmed: 0
- No difference: 13 — the signal didn't change the odds.
- Volatility illusion: 7 — bigger moves both ways, but no better direction.
- Worse than baseline: 4 — the signal did worse than similar stocks.
Results
* For the crypto rule, a hit means +10% within 24 hours on Upbit KRW markets. Each result links to an English detail page and a Korean write-up.
Limitations
- About one year of data, mostly one market regime. These rules may behave differently in other periods.
- "Up after 20 days" is one lens. It is not a full trading backtest with stops or position sizing.
- This is not financial advice, and it does not prove technical analysis is useless. It only shows that these one-line rules, on their own, did not beat a fair baseline here.
Past data · not investment advice