Our graveyard is longer than our portfolio.
Every idea below was tested with the same pipeline as our live strategies, and failed. We keep the list public because it shows what our filters actually reject.
| Strategy | Market | Failed at | Why |
|---|---|---|---|
| SuperTrend scalping | Indices, M5 / M15 | Generation | Nine hours of search, zero candidates accepted. |
| Larry Williams 9.1 and 9.2 | Indices | Out-of-sample | Published rules, no edge on unseen years. |
| Machine-generated systems | Silver, H1 | Re-tests | 273,000 generated, 22 passed the first filter, 0 survived the re-tests. |
| Machine-generated systems | S&P 500, H1 | Full-history test | Every finalist went flat from 2012 to 2019. Profit only after 2019. |
| Machine-generated systems | Dow Jones, H1 | Full-history test | Most finalists stalled from 2013 to 2016. |
| Day-trading systems | DAX, H1 | Out-of-sample | All finalists made money only from 2014 to 2019. |
| Machine-generated systems | USD/JPY | Validation | 73 candidates, none passed. |
| Machine-generated systems | EUR/USD, H1 | Generation | Nine hours, zero candidates accepted. |
| Bollinger Squeeze | Gold | Monte Carlo | Collapsed in the worst simulated histories. Removed from the portfolio. |
| Opening-range breakout from a popular video | Indices | Out-of-sample | Worked only from 2022 to 2026. |
| RSI 20/80 with 20-period average | Gold, M15 | Out-of-sample | Worked only from 2022 to 2026. |
| Elephant Bar | Indices | Out-of-sample | No edge after costs. |
| Gap fill | US indices | Backtest | No edge after costs. |
| Overnight mean reversion | US indices | Backtest | No edge after costs. |
| Turn of the month | S&P 500, Dow Jones | Backtest | No edge after costs. |
| Short-side mean reversion (RSI 2, Double Seven) | Indices, gold | Backtest | The long side works; the short side does not. |
| Short-term mean reversion | Oil, silver, USD pairs, AUD/NZD | Backtest | Reversion works in indices, gold and similar-economy crosses, not here. |
| Moving-average crossover | S&P 500, Dow Jones, oil | Backtest | No stable edge. |
| Asian range breakout | FX, gold | Backtest | No edge after costs. |
| Previous-day high/low sweep | FX, gold | Backtest | No edge after costs. |
| Bollinger + ADX reversal | FX | Backtest | No edge after costs. |
| Keltner pullback | FX | Backtest | No edge after costs. |
| Stochastic Pop | Indices | Backtest | No edge after costs. |
| Index cointegration pairs | US indices | Backtest | Relationship not stable enough to trade. |
| CFTC positioning (COT) filter | Gold, GBP/JPY | Portfolio fit | Did not improve the existing strategies. |
Tests run between August and October 2026. Market and timeframe listed where relevant.
What the graveyard teaches.
Short-term mean reversion works in stock indices, gold and currency crosses of similar economies. It fails in oil, silver and most USD pairs. A strategy that only works in the last few years is almost always a strategy that only worked in the last few years.