Oil-Signal Backtests: Another Round of Humbling Results
There is something seductive about an oil-price signal that promises to time energy stocks. Brent closes below its 20-day moving average, so you buy a refiner. XOM lags CVX badly, so you bet on a catch-up. Oil leads the complex, or maybe it doesn't. The platform's latest run of backtests puts several of these instincts to the test — and the results are consistently humbling.
The Intraday Close Effect That Wasn't
Take the idea that after a high-volume, Brent-led selloff, the way XLE closes within its intraday range tells you something about the next five days. The intuition is clean: a close near the lows suggests panic, a close near the highs suggests resilience. The data, however, says no. Across 67 high-volume Brent-led selling days, XLE's forward 5-day return averaged 1.04% when it closed in the top half of its range and 0.91% when it closed in the bottom half. That gap of 0.13 percentage points is indistinguishable from noise, with a Welch p-value of 0.87. Medians told the same story (2.13% vs 1.24%). The close-location effect, in other words, is a mirage.
A Sweep of Backtests, All Trailing the S&P
The bigger picture emerges from four backtests on energy tickers, all run against the same benchmark window. Buy MPC when Brent closes below its 20-day MA? The strategy returned +9.26% on $100,000 across 38 trades, but with a 37% win rate and SPY returning +68.30% over the same window, it trailed the benchmark by 59.05 points. Buy MPC when its 5-day total return underperforms Brent's? That did better — +28.03% across 30 trades, a 60% win rate — but still trailed SPY by 40.27 points. Buy XOM when its 5-day move underperforms CVX's by more than 2%? +25.08% across 31 trades, a 61% win rate, and a 43.22-point shortfall. And buying SLB when Brent closes above its 20-day MA? That lost money outright: -16.69% across 34 trades, trailing SPY by 85.00 points.
None of these strategies are crazy. They all have a ratio, a logic, and enough winners to feel real. Yet every single one would have been better off simply holding the S&P 500. The best absolute return, the MPC relative-strength trade, still underperformed by a wide margin. The only consistent winner in this set appears to be the benchmark.
No Lead-Lag, No Regime Edge
Finally, consider the XOM-Brent relationship. A common narrative is that oil majors move first when crude is below its 50-day moving average, and Brent leads when crude is above. The platform tested exactly that, looking at 407 below-SMA days. XOM's trailing 5-day return predicted Brent's forward 5-day return with a slope of -0.013 and a p-value of 0.934. Brent trailing XOM was equally dead, with a slope of -0.021 and p=0.734. In plain terms: neither series tells you anything about the other's next move. The regime doesn't change that.
What ties all these results together is a simple lesson for quantitative research: energy markets are crowded with narrative-driven participants, and the signals that sound smart in conversation tend to be exhausted or just noise by the time they are coded up. The platform's value here is not in finding a hidden edge — it's in publishing the failed tests so that the rest of us stop chasing them. The data leans toward humility. Tactical oil-signal trading, at least as defined by these rules, has produced nothing that justifies abandoning a buy-and-hold posture. That is a finding worth more than a winning backtest.