Triple RSI Strategy

Triple RSI buys short-term panic in an otherwise healthy market. It watches a fast 5-period RSI and only acts when four conditions line up at once — three RSI conditions plus a trend filter — then exits when RSI mean-reverts. It runs on SPY as a market-timing signal (/api/strategies) and, via the scanner, on individual stocks (/api/scan).

• Buy when all four entry conditions are true (see §3).
• Sell when RSI(5) crosses back above 50.
• Typical hold: a few trading days (mean ~5; range 1–15 in the backtest).

Best for

• Choppy-to-rising, mean-reverting markets that hold above their 200-day average and shake out weak hands on brief dips.

Avoid when

• Strong, sustained downtrends and momentum crashes, where "oversold" keeps getting more oversold. The 200-day filter reduces but does not eliminate this.

Market behaviour exploited

Short-term overreaction (mean reversion) inside a primary uptrend — the tendency of healthy markets to snap back after a fast, shallow pullback.

The edge is a tendency, not a guarantee; mean reversion fails hard in trending bear markets, which is why the exit and the trend filter matter.

One pure core (tripleRsiSignalFromBars) backs the live SPY state, the scanner, and the backtest adapter, so the maths lives in exactly one place. Parameters (TRIPLE_RSI_PARAMS): RSI period 5, SMA period 200.

Entry — all four must be true

1. RSI(5) < 30 — deeply oversold.
2. RSI(5) falling for 3 consecutive sessions — momentum still washing out, not already turning.
3. RSI(5) was < 60 exactly 3 sessions ago — the drop began from a non-overbought state (not a blow-off top).
4. Close > 200-day SMA — the primary trend is still up.

Exit

RSI(5) crosses above 50 — the mean reversion is complete.

Position management

• The rules define no fixed price stop or profit target — the exit is condition-based (RSI reclaims 50). So stopLoss, firstTarget, riskPercent, and rewardRiskRatio are reported as null (never invented). entryPrice is the signal-bar close.
• Risk is controlled by only entering above the 200-day trend and by the mean-reversion exit.

Scanner

The identical per-symbol logic runs across a universe via GET /api/scan?strategy=triple_rsi, returning each match's signal, a 0–100 confidence score, the passed-rule checklist, and metrics (rsi5, sma200, close).

Deterministic, from the shared buildConfidence framework. confidenceScore is a setup-quality measure (0–100), not a probability of profit. Components (maxima sum to 100):

A bare BUY (all four gates met, RSI just under 30) scores ≈ 70 (Good); the more deeply oversold, the closer to 100 (Exceptional). On /api/strategies the live SPY score is null when no setup is active; on each /api/scan match it is always populated. Bands: <70 Weak · 70–79 Good · 80–89 Strong · 90–100 Exceptional.

The API returns the labels of the rules that currently pass; a full BUY shows all four:

• RSI(5) below 30 (deeply oversold)
• 3-day RSI decline (momentum still falling)
• RSI was below 60 three days ago (came from a non-overbought state)
• Above 200-day MA (primary trend up)
• Enter at the signal-bar close; no fixed stop/target — exit is RSI(5) reclaiming 50.

• High hit-rate, positive expectancy. 89.9% win rate and a 6.21 profit factor over 89 real trades and 30 years.
• Shallow drawdown. 7.4% max drawdown — the trend filter and quick exit keep losers small (worst trade −7.37%).
• Objective and mechanical. Four unambiguous numeric gates; no chart-reading or discretion.
• Dual use. The same core powers a SPY timing signal and a full stock scanner.
• Short exposure. In the market only a few days per trade, reducing overnight/regime risk.

• No hard stop. The exit is condition-based (RSI reclaims 50); a trade can run against you until it does. Worst trade was −7.37%.
• Regime-dependent. Mean reversion fails in sustained downtrends; the 200-day filter mitigates but doesn't remove this.
• Selective / bursty. Only ~3 trades/year on SPY, clustered in volatile stretches — long quiet periods with no signal.
• Close-to-close model. No commissions/slippage modelled; the backtest fills at the signal-bar close.
• Small average edge. ~1.3% per trade — sensitive to costs and execution.

• Static snapshot. historical.* is baked as of 2026; the live /api/backtest always recomputes from current data.
• Single-vendor, close-to-close. Prices from Yahoo; SPY history starts 1993-01-29; no slippage/commissions modelled.
• **No historicalPattern.** Deliberate — the holding window is variable, so a fixed-offset average would misrepresent the strategy.
• Backtested on SPY. The scanner applies the same rules per-symbol across a universe, but the published statistics are the SPY single-symbol backtest.