Thesis
RSI Snap-Back bets that large-cap tech names overshoot on short-term momentum and then revert. The mechanics are simple and disciplined: buy the most oversold names when RSI drops below 35, exit when RSI pushes above 70, and hold no more than four positions at once. That hard 4-slot book is the strategy's main risk control — it caps how much concurrent drawdown exposure the book can accumulate and forces rotation into the freshest oversold signals across a tight universe of AAPL, MSFT, NVDA, GOOGL, AMZN, META, and TSLA.
Backtest performance
Over 451 days, the backtest returned 20.95%, growing a $10,000 book to $12,095. That works out to a CAGR of 11.21%, with a 66.67% win rate across 37 trades. Trading costs were negligible — $37 in fees, no FX — which is unsurprising given only 37 fills, though turnover ran hot at 773%, reflecting the rotate-and-replace design.
The risk picture is more mixed. Sharpe sits at 0.61, respectable but not commanding, and max drawdown reached 23.73% — nearly the full magnitude of the total return. In other words, an investor riding this strategy would have needed the stomach to watch almost a quarter of peak equity evaporate before the reversion thesis paid off. That is the trade-off inherent to buying falling knives: the entries are, by construction, into weakness.
Recent activity
Here is the caution flag. The last six scheduled runs — from August 10 through August 17 — each report 0 trades executed, 0 rejected, with the book holding $10,000 in cash and $10,000 total. The strategy is nominally live, but in practice it has been fully idle, parked in cash with no open positions. No name in the universe has printed an RSI below 35 recently enough to trigger an entry, so the model is simply waiting.
That is not a malfunction — a mean-reversion system should stand aside when nothing is oversold — but it does mean the live track record is currently empty. The headline metrics are backtest figures, not realized live performance.
Risks and read
Two things temper the strong backtest. First, validation is null — there is no recorded walk-forward or out-of-sample check, so we cannot yet distinguish genuine edge from curve-fit on a 451-day window. Second, the strategy leans entirely on seven correlated mega-caps; a broad tech drawdown could push several names oversold at once, filling all four slots into the same macro shock rather than diversifying the reversion bets.
The verdict: an intuitive, low-cost, disciplined design with a plausible edge and a believable win rate — but one whose drawdown rivals its return, whose live book is presently flat, and whose out-of-sample robustness remains unproven. Worth watching once it starts trading live; not yet worth trusting on the backtest alone.