Is Patrick Boyle’s strategy legit? We backtested it.
The claim: “Sell in May and go away?”
Main test: S&P 500 (SPY daily, 1993-2026). Every test is in the table below.
What we found
- On the S&P, Nov-Apr beat May-Oct (6.8% vs 4.3% a half-year), but holding all year still made more (10.7% vs 6.6% a year).
- No effect on the Nasdaq.
Full test notes, including earlier and shorter tests
- Nasdaq: no effect (5.8% vs 6.4%).
- Bitcoin since 2021: Nov-Apr +26% vs +2%, only 5 winters.
Every test we ran
| Strategy | Market | Trades | Win rate | After costs | Versions profitable | Verdict |
|---|---|---|---|---|---|---|
| Sell in May | S&P 500 (SPY daily, 1993-2026) | 33 | — | profitable | 0/4 | Weak edge |
| Sell in May | Nasdaq 100 (QQQ daily, 1999-2026) | 27 | — | — | 0/4 | Hype |
R = multiples of the risk taken on each trade (−1R = one full stop-loss). Results include trading costs.
How we tested it
We turned the rules into code, ran them on years of real prices with trading costs, then re-ran nearby versions of the rules so one lucky setting can’t pass.
The exact entry, stop-loss, exit and filters behind every number on this page.
Monthly SPY/QQQ with dividends (Yahoo), Bitcoin 2021-2026 (Dukascopy).
0.02% per switch
4 versions tested - the full list with settings is in the paid report.
See every version ▾Nearby versions: why one lucky setting can’t pass
For every test we also ran nearby versions of the same rules: different profit targets, stop-losses, time windows, entries or filters. A strategy only counts as real when most of its versions make money after costs, not just the one that was shown in the video.
■ 75%+ of versions profitable ■ 40–74% ■ under 40%
Other channels we tested
in a $10 reportGet the rules →
We test the rules, not the people. Backtest results only; not financial advice. Past results do not guarantee future results.