Thesis
RSI Snap-Back is a mean-reversion play on large-cap tech. The idea is simple and well-worn: the Mag-7 names snap back sharply after short-term momentum extremes, so the strategy buys the most oversold names (RSI below 35) and exits the overbought ones (RSI above 70), rotating to hold a tight four-name book. The hard four-slot cap is the discipline mechanism — it limits how much concurrent drawdown exposure the strategy can take on at any one time. The trading universe is AAPL, MSFT, NVDA, GOOGL, AMZN, META, and TSLA.
Recent Activity
Here the story is quiet to the point of inert. Across six scheduled runs spanning 2026-09-30 through 2026-10-08, the live strategy executed zero trades and rejected zero candidates. Cash and total equity have held flat at $10,000 the entire window. In plain terms, nothing in the universe has dipped to the RSI-below-35 entry threshold, so the book has stayed fully in cash.
That isn't a malfunction — it's the strategy behaving as designed. A mean-reversion system that only fires at momentum extremes should do nothing when none of its names are oversold. But it does mean there is no fresh live P&L to evaluate, and the live track record remains effectively untested against real conditions.
Backtest & Validation
The backtest covers 451 days and 37 trades, ending at $12,095.14 in equity — a total return of 20.95% and a CAGR of 11.21%. The win rate is a healthy 66.67%, and trading costs are negligible: $37 in total fees and zero FX cost.
The numbers that give me pause sit next to those headlines. The Sharpe ratio is 0.61 — modest, indicating the returns came with meaningful volatility rather than a smooth ride. More striking, the maximum drawdown is 23.73%, slightly larger than the entire total return. A strategy whose worst peak-to-trough fall exceeds its full-period gain is one you have to be able to stomach on the way up. Turnover of 773.36% also tells you this is an active rotator, not a buy-and-hold — though the tiny fee total confirms the churn hasn't been expensive.
The most important field is the one that's empty: validation is null. There is no out-of-sample or walk-forward validation on record here. For a mean-reversion rule fitted to a specific, trending basket of seven names, that's the key open risk — a 20.95% in-sample return tells us little about robustness until it's confirmed on data the rules never saw.
Verdict
RSI Snap-Back is a clean, disciplined concept with a solid in-sample win rate and trivial costs. The strengths are real; so are the caveats. The drawdown profile demands conviction, the Sharpe is unremarkable, and without validation the edge is unproven. For now it's a patient book — waiting for an oversold signal that simply hasn't come.