The thesis
The mean-reversion strategy runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it enters when the 14-period RSI drops below 30 (oversold) and exits when RSI climbs above 70 (overbought). It trades a universe of 24 large-cap US names — mega-cap tech, financials, staples, healthcare and energy (AAPL, MSFT, NVDA, JPM, JNJ, XOM and peers). This is a sensible sandbox for the approach: liquid, well-behaved names where short-term dislocations tend to snap back.
Recent activity
The live paper account has been strikingly inactive. The scheduled runs from 7–14 September all report the same line: 0 executed, 0 rejected, cash steady at $7,569.25 and total equity drifting between roughly $9,788 and $9,872. The last actual fill was a single buy — 21 shares of WMT at $115.75 on 31 May 2026 — and nothing has traded since.
That quiet is the thesis working as designed, not a malfunction: RSI simply hasn't crossed its thresholds across a mostly calm, grinding tape. But it carries a cost. With most of the book in cash and one legacy position, total equity has been sitting below the $10k starting mark for weeks. A strategy that only acts on extremes will, by construction, spend long stretches doing nothing — and give up ground when markets rise without ever getting oversold.
Backtest and validation
On paper the full-sample backtest looks appealing: +14.73% total return over 451 days (7.98% CAGR), a 70.59% win rate across 38 trades, and a modest 15.64% max drawdown. Fees were negligible ($38 total). The high win rate is the classic mean-reversion signature — many small, reliable gains.
The problem is what happens under scrutiny. Validation failed. Walk-forward testing over four folds shows three positive and one negative, but the decisive out-of-sample fold (Dec 2025–May 2026) returned −2.84% with a −0.33 Sharpe and a 14.96% drawdown — nearly the whole full-sample drawdown concentrated in the most recent, unseen period. The deflated Sharpe ratio (DSR) of just 0.304, against 6 trials, says the headline Sharpe of 0.58 is largely explained by selection rather than genuine edge. PSR of 0.785 is decent but not enough to rescue the verdict.
Strengths and risks
Strengths: a genuinely high hit rate, low trading costs, disciplined rules, and honest folds — two-thirds of the backtest window was positive, and the early-2025 fold posted a strong 1.32 Sharpe.
Risks: the edge does not survive out-of-sample, the modest Sharpe is deflated once trial count is accounted for, and the recent dormancy plus sub-$10k equity suggest the signal is misfiring in the current regime. Mean reversion also carries tail risk — it buys falling assets, so a trending sell-off can turn many small wins into one large loss.
Verdict
Keep it live for observation, but treat the backtest number with caution. A 70% win rate is seductive; a failing validation and a negative out-of-sample fold are the more honest signals. Until the strategy demonstrates edge on data it hasn't seen, it belongs in the watch-and-learn column — not the capital-allocation one.