The thesis
RSI Snap-Back is a mean-reversion strategy trading the Magnificent Seven: AAPL, MSFT, NVDA, GOOGL, AMZN, META and TSLA. The logic is deliberately plain — large-cap tech names tend to rebound after short-term momentum extremes, so the strategy buys the most oversold names (RSI below 35) and sells into overbought strength (RSI above 70). A hard four-slot book caps concurrent exposure and forces the strategy to rotate rather than sprawl. It's a textbook contrarian setup on a universe liquid enough to make the reversion assumption plausible, and the slot limit is a genuine risk control rather than decoration.
Recent activity
The last two executed trades were a buy of 10 NVDA at $204.04 on 2 November and 7 GOOGL at $280.28 on 30 October. Since then, six consecutive scheduled runs (3–10 November) each closed with 0 executed, 0 rejected — cash pinned at $5,956.92 the entire time. That is the discipline working as designed: with no name breaching the RSI < 35 trigger, the strategy simply sits. But it also means only two of four slots are filled and roughly 62% of the book is in cash. Total equity drifted down across that quiet week, from $10,022.55 on the 4th to $9,637.31 on the 10th — the strategy is now underwater on its starting stake while waiting for a signal that hasn't come.
Backtest and validation
Over a 451-day backtest the strategy returned 20.95% (CAGR 11.21%), ending at $12,095.14 on a win rate of 66.67% across 37 trades. Fees were trivial ($37 total, no FX cost). Those are respectable headline numbers. The caveats are real, though:
- Sharpe of 0.61 is modest — the returns came with meaningful volatility, not smooth compounding.
- Max drawdown of 23.73% actually exceeds the total return, so an investor could have sat through a deeper hole than the final gain.
- Turnover of 773% confirms this is an active rotator; the thin fee drag held up in backtest, but it leaves less margin if spreads or slippage bite in live trading.
Most importantly, the validation field is null — there is no separate out-of-sample check on record, and the live metrics mirror the backtest exactly. Until live results diverge meaningfully from the fitted curve, the 20.95% should be read as in-sample promise, not proof.
Verdict
RSI Snap-Back is coherent, well-governed and cheap to run, with a win rate that supports its reversion premise. The open questions are whether that edge survives out of sample, whether the 24% drawdown is tolerable, and whether the current half-cash, two-name posture is patience or a thesis quietly starved of signals. Worth watching — with the validation gap front of mind.