Backtest: Buy MPC at the close when its 5-day total return underperforms Brent crude's...
A trading rule that wins 60% of its bets can still come up well short of the market. This one buys Marathon Petroleum when its five-day return lags Brent crude by more than five percentage points, then exits on a snap-back signal or after ten sessions. Across 30 closed trades, it returned 28.03% on $100,000 in starting capital — and still trailed a buy-and-hold SPY position by 40.27 points over the same window.
The thesis is clean: a refiner underperforming crude that sharply usually signals a crack-spread shock, and those dislocations tend to reprice. The real question is whether that reversion is worth harvesting on its own. The full backtest below walks through every trigger, every exit, and the trade-by-trade evidence behind that benchmark gap.
Buy MPC at the close when its 5-day total return underperforms Brent crude's 5-day return by more than 5 percentage points; exit after 10 trading days or when MPC's 5-day return outperforms Brent's by 2 percentage points, whichever comes first. Extreme refiner underperformance versus crude marks a crack-spread shock that reprices the crude move and tends to snap back.
How this was measured
This is a simulated backtest generated from the plain-English strategy below, executed bar-by-bar on historical market data using the price + news data mode with $100,000 starting capital. Strategy: Buy MPC at the close when its 5-day total return underperforms Brent crude's 5-day return by more than 5 percentage points; exit after 10 trading days or when MPC's 5-day return outperforms Brent's by 2 percentage points, whichever comes first. Extreme refiner underperformance versus crude marks a crack-spread shock that reprices the crude move and tends to snap back.
The key numbers
The charts
The takeaway
The strategy returned +28.03% on $100,000 starting capital across 30 closed trades with a 60% win rate. Over the same window SPY buy-and-hold returned +68.30%, so the strategy finished trailing the benchmark by 40.27 points. Best single trade +17.64%, worst -13.30%.
The fine print
- Simulated results on historical data — fills, slippage and costs are idealized.
- Past performance does not predict future results.