Retired: 2026-07-12
53.5%
break-even 48.5% at its realized average win/loss
re-measured (top 10 coins · 1H · 2026-10-06): 50.36%
Percentage of trades that were profitable
1.22
re-measured (top 10 coins · 1H · 2026-10-06): 0.83
Gross profit / gross loss ratio
as published (50 coins · 12H · 2026-05-04)
Marks next to a metric: ~ = borderline (PF between break-even and 1.5; win rate ≤5 pts above break-even, or 50–55% if none shown) · ! = below the bar, or the strategy is already retired or shelved — a closed verdict outranks its number. No mark = clears it and is still standing.
Wins only under specific regimes or windows in our measurements — the card's bull/bear PF row shows the regime side; strategies that win both windows were still killed by half-year walk-forward instability. A past condition profile, not a recommendation — conditions earn trust only by surviving forward.
Card re-measured: 2026-10-06 · 3,862 trades analyzed
Overview
The Stochastic RSI Short strategy applies the Stochastic oscillator to RSI values rather than price. By measuring whether RSI itself is in an “overbought of overbought” condition, the indicator catches trend exhaustion moments that raw RSI or price-based Stochastic alone often miss.
Discovered 2026-05-04 via multi-window parameter sweep across 50 coins on a 12H timeframe. The strategy produced 2,674 total trades — one of the highest sample sizes in the current strategy library — giving statistical confidence in the measured profit factor of 1.22.
How It Works
- Stochastic RSI calculation — RSI(14) is calculated, then the Stochastic oscillator (3-period smoothing) is applied to that RSI, producing %K and %D lines ranging 0–100
- Overbought zone — %K must be above 80 (overbought territory; %D has no threshold check)
- Death cross — %K crosses below %D in the overbought zone, signaling momentum exhaustion
- Entry — SHORT on the next bar’s open after the cross is confirmed
- Exit — TP 15% / SL 15% / 4-bar (2-day on 12H) timeout
Why It Works (Thesis)
The Stochastic RSI death cross in overbought territory captures a specific market condition: the rate of change of momentum is reversing, even as the market may still be rising. By the time %K crosses %D above 80, the underlying RSI has been at extreme levels for multiple periods and is now losing steam — a reliable sign that short-term sellers are entering.
The 12H timeframe is key. At shorter timeframes, Stochastic RSI signals fire constantly and are mostly noise. At 12H, each signal represents a half-day of sustained overbought momentum followed by a measurable cross — far more significant than hourly whipsaws.
The symmetric SL and TP (both 15%) reflects the strategy’s nature: it captures mean-reverting moves of moderate size, not home-run directional trends.
Results (2-year backtest, IS/OOS split, measured 2026-05-04)
| Metric | IS (May24–May25) | OOS (May25–May26) | Combined |
|---|---|---|---|
| Total trades | 1,434 | 1,240 | 2,674 |
| Win rate | ~53% | ~54% | 53.5% |
| Profit factor | 1.07 | 1.48 | 1.22 |
| Coins profitable | 27/50 | 37/50 | — |
OOS/IS ratio: 1.38 — OOS profit factor (1.48) exceeded IS (1.07) by 38%. The OOS period (May 2025–May 2026) was primarily bearish, which favors SHORT strategies — explaining why OOS outperformed IS.
The IS period (May 2024–May 2025) included the 2024 bull run where SHORT signals naturally underperform. The IS PF of 1.07 is profitable but modest; the strategy showed its best edge when market conditions aligned (bear regime).
Caveats
- IS PF of 1.07 is marginal — in bull markets, this strategy produces near-flat results. It is strongest in bear/sideways regimes.
- 53.5% WR is below the ≥55% threshold we prefer for verified strategies; the edge comes from win/loss size balance at symmetric 15%/15% rather than win rate.
- 2,674 trades is excellent for statistical confidence but means the strategy fires frequently — monitor for signal fatigue if deploying on multiple coins simultaneously.
- 12H signals average about one per coin every 14 days (2,674 trades / 50 coins / 2 years).
- Not live-tracked on OKX. Backtest only.
Update (2026-08-08): retired
The numbers above are the original run (50 coins, 12H, 2,674 trades). A current measurement on the simulator’s universe reversed the sign, and this preset is now published in the failures section of the grid:
| Original | Current (2026-08-08) | |
|---|---|---|
| Profit factor | 1.22 | 0.86 |
| Win rate | 53.5% | 51.0% |
| Total return | — | -496.3% |
| Sharpe | — | -1.20 |
| Trades | 2,674 | 3,643 |
Source: src/config/simulator-presets.ts (stochastic-rsi, measuredAt: 2026-08-08).
The original numbers stay as written — the point is the gap, not a tidier page.
Re-measurement note (2026-08-12): we could not reproduce the 2026-08-08 kill-run numbers on the current engine, which received a drawdown-accounting repair on 2026-08-11 (pre-repair, this strategy’s equity excursions ran uncapped past −100%). Post-repair re-runs on the same universe: registry defaults (SL 15 / TP 15) → PF 1.25, +55.2%, Sharpe 1.10; preset defaults (SL 10 / TP 5) → PF 1.04, +6.1%, Sharpe 0.24. Neither shows the sign-flip. The kill-run figures they replace were PF 0.86, total return -496.3%, Sharpe -1.20 — a total return past -100% is the signature of the pre-repair drawdown bug, which is why they are quoted here with their cause and not in the card blurb. The retired status stays until this discrepancy is reviewed — evidence that stopped reproducing needs review, not a silent un-kill.
Why the card still reads PF 1.22 (2026-05-04): the 2026-08-12 re-run reports PF, total return and Sharpe, but not trade count or win rate. Replacing only the PF would mix two runs in one row — 1.25 from the repaired engine beside 2,674 trades and 53.5% from the original. The card therefore keeps the 2026-05-04 row whole and states its date; this note holds the newer number with its scope.