The thesis
RSI Snap-Back is a straightforward mean-reversion play on large-cap tech. The premise: the Mag-7 names — AAPL, MSFT, NVDA, GOOGL, AMZN, META, and TSLA — tend to snap back sharply after short-term momentum extremes. The strategy buys the most oversold names on RSI below 35, exits into strength above RSI 70, and rotates to hold no more than four names at once. That hard four-slot cap is the risk-management spine of the design: it enforces selectivity and limits how much concurrent drawdown the book can accumulate.
It's an opinionated but well-worn idea. Reversion works best in liquid, heavily-traded names where short-term dislocations get arbitraged away quickly — and the Mag-7 fits that profile as well as any universe on the market.
Backtest performance
Over a 451-day backtest, the strategy returned 20.95% (final equity $12,095 on a $10,000 base), an 11.21% CAGR. The win rate is a respectable 66.67% across 37 trades, and fees were negligible at $37 total with no FX cost.
The less flattering numbers sit alongside them. The Sharpe of 0.61 is modest — this is a strategy that earned its return by accepting real volatility, not by being smooth. The max drawdown of 23.73% is steep for a book that caps itself at four names, and it slightly exceeds the total return, which tells you the equity curve was a bumpy ride. Turnover of 773% reflects the rotate-and-rebalance mechanic; it's manageable here only because modelled fees are tiny, and it would deserve closer scrutiny under a more realistic cost regime.
Recent activity
Here's the part worth flagging. Across the last six scheduled runs — August 20 through 27 — the strategy executed zero trades every single day and has remained entirely in cash at $10,000. No RSI reading in the universe has dipped below the 35 entry threshold, so the book has simply stood down.
That's the system working as designed: no signal, no trade. Discipline over activity. But it also means the live track record is currently a flat line, and none of the backtested edge is being expressed right now. A reversion strategy needs dislocations to feed on, and the Mag-7 hasn't handed it any recently.
Risks and the missing piece
The headline caveat is that validation data is null — there's no out-of-sample or walk-forward result attached to this strategy. A 20.95% in-sample backtest with a 0.61 Sharpe and a near-24% drawdown is suggestive, not conclusive, and without validation we can't distinguish genuine edge from curve-fit to one favourable window.
Combined with the sub-1 Sharpe and the drawdown that swallows the return, the honest read is: a sensible, disciplined design with a plausible thesis, currently unproven out-of-sample and currently inactive. Worth watching for how it behaves when the next tech dislocation actually arrives — that's when we'll learn whether the snap-back is real.