The 200-day + RSI(2) pullback on the 400 most liquid stocks: does it still work after costs?
Buying a stock above its 200-day average after its 2-day RSI drops under 5 (and shorting the mirror image), selling on the bounce: does it pay per trade after costs on the 400 most liquid US stocks, including names that later delisted?
No. Neither side passed. The long side won 63.5% of 8,402 trades and made +0.28% a trade after costs, but losses were about 1.4 times the wins and the month-clustered lower bound sat below zero. The short side made +0.58% a trade on 4,418 trades but lost money in 3 of 6 years and missed the deflated-Sharpe bar. Run as a 10-stock book, the long rule made 3.9% a year in 2016 to mid-2022 while SPY made 12.0%. The holdout stays sealed.
Backtested by QS Research
Pre-registration
Two trials fixed before any return was read: long (close above the 200-day average, RSI(2) under 5) and the short mirror, each exiting on the bounce. The 400 most liquid US stocks on each day, $5 and up, listed and delisted names, 2016 to 2026, split 60/20/20 by signal date; judged at alpha/2 with a month-clustered bound, a year check, a deflated Sharpe at an effective N of 4 and an SPY-relative check. Costs are the spread by liquidity tier both sides and, for shorts, borrow. One amendment before any return: the data pins, stage boundaries and the rule-only signal table.
Registered 2026-11-15Amendments
- 2026-11-15 Data pins, stage boundaries, the rule-only signal table and the rule readings, recorded before any return was read. approved: Recorded before the counted run
A holdout sample is sealed: its results stay unread until the study is final.
Trials · 2 trials · 0 passed · judged at N = 2 trials; 2035 trials run in total
| Trial | Verdict | |
|---|---|---|
| mr2-long | FAIL | counted |
| mr2-short | FAIL | counted |
Limitations
- The stock, not the option: the cards offer a call or a put, which also pays decay and its spread.
- Daily bars: fills at the official open, with no capacity limit.
- No stop; delisted names exit at their last close.
- Discovery was not fresh for the long side: earlier, survivorship-biased looks at this rule are disclosed in the pre-registration.
- No ETFs, no options-volume filter and no buffer in the judged universe.
- Holdout sealed: the years from September 2024 were never read.
The numbers: every trial's checks, metrics, curve and yearly results, and the provenance
1 published version · current 1ba724b9be6924fe, published 2026-11-15