Energy Backtests Win Often and Still Lose to the Index

The energy tape's most telling number this week isn't a barrel price. It's a rounding error: Valero trading after-hours at EUR 342.00, up 0.04 percent against its prior close. Here is a refiner with a Q3 report date on the calendar, sitting in a sector that generates constant noise, and it moved essentially nowhere. The headlines tagged to Exxon Mobil this week weren't even about Exxon — a Form 4 for Airbnb, a 52-week low for Liberty Global, an SEC ruling on a Tesla-crafted proxy-voting plan. That kind of loose tagging is its own signal about how much genuine energy-specific information is flowing.

High win rates, small dollars

The platform's own backtests this week tell a remarkably consistent story about how energy mean-reversion actually behaves. Three setups, one pattern: the hit rate is gorgeous, the dollars are not.

Buying BP at the close after a 1% down day ran 5 closed trades at an 80% win rate and returned +1.25% on $100,000. The BKR variant — buying at a 20-day closing low with Brent in the frame — ran 12 trades, won 91% of them, and returned +6.95%. The VLO/USO ratio setup ran 8 trades at 75% and returned +16.75%. Every one of those trails SPY buy-and-hold's +85.28% over the same window, by 84.03, 78.32 and 68.53 points respectively.

The takeaway isn't that these strategies are broken. It's that win rate is a comfort blanket. A 91% hit rate paired with a worst trade of -5.64% and a best of +6.31% describes a distribution that pays small and often while quietly accumulating the occasional real loss. Five trades, eight trades, twelve trades — those sample sizes are thin enough that a single outlier flips the story. Traders love to quote hit rates. The data here leans hard the other way: frequency is not edge.

Oil's plumbing, not its direction

The more durable finding this week concerns USO's tracking error to Brent. When Brent's realized volatility sits in the top quintile — averaging about 60.5% annualized — USO's trailing 20-day tracking error averaged 44.2% annualized. In the calm bottom quintile, where Brent RV sits near 20.7%, that tracking error averaged just 15.8%. That's a 28.4-point gap and a 2.79x ratio.

The implication is uncomfortable. In precisely the moments when an oil proxy is most in demand — volatile, headline-driven sessions — it tracks its benchmark worst. Anyone treating USO as a clean expression of a Brent view is carrying a second, volatility-conditional bet they may never have priced. That's a structural fact about the wrapper, not a forecast about crude.

The headline trade is a coin flip

The supply-shock intuition looks shakier too. On the idea that refiners lag on supply-shock headline days, the evidence refuses to cooperate: across 61 top-quintile headline sessions, VLO's 5-day forward return minus XLE's averaged +0.95%, versus +0.66% across 395 low-headline days. The 0.28-point gap points the opposite direction from the thesis, and sits within coin-flip territory.

The XOM-versus-XOP setup is the more intriguing one. When unemployment ticks up month-over-month and Brent's 20-day return is negative, the integrated name averaged +0.68% over the following 20 sessions while the E&P proxy lost 1.67% — a 2.34% edge, across 9 clean, non-overlapping instances. Suggestive, not proven. Nine observations is a hypothesis, not a conclusion.

So what is the tape actually saying? Eni rolling out a 30% power and gas discount and TETRA scheduling Q3 results for November 2 are consumer-facing and calendar-facing items — useful context, weak directional signal. Meanwhile the research keeps returning the same verdict: energy's tradable edges are conditional, small, and easily mistaken for edge when they're merely frequent. The mood around refiners may be mildly constructive. The data keeps saying the interesting question isn't which way oil goes — it's what you're actually holding when you try to express that view.