The thesis
RSI Snap-Back is a mean-reversion play on the seven largest US tech names — AAPL, MSFT, NVDA, GOOGL, AMZN, META, and TSLA. The logic is simple and disciplined: buy the most oversold names when their RSI drops below 35, exit into strength when RSI climbs above 70, and never hold more than four positions at once. That hard 4-slot book is the strategy's risk governor, capping concurrent exposure and forcing rotation rather than sprawl.
The bet is that large-cap tech, even in a secular uptrend, overshoots on short-term momentum extremes and snaps back. It's a well-worn premise, and on this universe it has a plausible edge.
Recent activity
The most striking thing about the last week is what didn't happen. Across six scheduled runs from September 17 through 24, the strategy executed zero trades and rejected zero — sitting flat in $10,000 of cash the entire time, total equity unchanged. No entries fired because none of the Mag-7 dipped under the RSI 35 threshold.
This is the discipline working as designed: no signal, no trade, no forced risk. But it's also a reminder that a mean-reversion book earns nothing while its universe grinds higher without pulling back. The strategy is currently a spectator, waiting for a dislocation that a strong tape may not deliver.
Backtest and validation
Over 451 days the backtest returned 20.95% (final equity $12,095), an 11.21% CAGR, across 37 trades with a 66.67% win rate. A two-thirds hit rate is genuinely respectable for a systematic strategy, and fees were negligible — $37 total, no FX cost.
The risk side is where enthusiasm should cool. The Sharpe of 0.61 is modest, meaning returns came with meaningful volatility rather than smooth compounding. More pointed: the max drawdown of 23.73% is nearly as large as the entire period's return. A return-to-drawdown ratio of roughly 0.9 tells you an investor had to stomach a peak-to-trough loss almost equal to everything the strategy made. Turnover of 773% also signals a busy, churn-heavy book — cheap here, but a real drag at higher fee tiers.
Critically, the validation field is null. There is no out-of-sample or walk-forward confirmation in this data. Every metric above describes the backtest and nothing beyond it, so the results should be read as in-sample until independent validation exists. For a mean-reversion rule tuned to a single, recently high-flying universe, that caveat is not a formality — it's the whole question of whether the edge survives contact with unseen data.
The verdict
RSI Snap-Back has an intuitive thesis, a healthy win rate, and admirable restraint — it would rather hold cash than force a bad entry. But the deep drawdown, unremarkable Sharpe, and complete absence of validation mean it hasn't yet earned unqualified confidence. The next thing worth watching isn't the return figure; it's a genuine out-of-sample test, and whether the book can put capital to work when the Mag-7 finally offers it a dip.