Oil's Favorite Shortcuts Don't Survive Contact with Data

The oil complex loves a good shortcut. Refiners are supposed to hedge crude spikes, oil majors are supposed to telegraph the next day's barrel price, and headlines about Hormuz are supposed to move everything. The latest batch of research from the trades.run quant pipeline suggests these shortcuts are more ritual than edge.

The great negative beta mirage

Start with refiners. A common piece of market lore is that refiners like MPC suffer when crude spikes because input costs jump before product prices adjust. The data over 37 extreme Brent-up days — moves above the 95th percentile, or up more than 3.9% — shows a beta of -0.16, which sounds like the expected dampening. But the 95% confidence interval runs from -0.50 to +0.18, and the p-value is 0.34. Statistically, that apparent hedge is indistinguishable from zero. The true relationship could just as easily be positive. One tail event can flip the sign.

Similarly, TTE's intraday range tells us something about supply-fear volatility. On 215 days when Brent rose more than 1%, the average (high-low)/close was 3.28%. On 218 down-crude days, it was 3.33%. That 0.05 percentage point gap points the wrong way relative to the fear hypothesis and has a t-stat of -0.17 (p=0.87). Up-day panic and down-day panic look identical in the intraday tape. The market's reaction to crude moves is not the clean, directional story traders tell themselves.

The leading indicator that barely leads

Oil traders often watch XOM as a bellwether for next-day crude. The research finds a statistically real relationship over 722 overlapping days: XOM's return on day t correlates with Brent's next-day return at 0.123, with a p-value around 0.0009. That's unlikely to be luck. But the regression slope is the sobering part — a 1% move in XOM is associated with only about a 0.22% move in Brent the next day. So yes, there's a signal, but it's a weak one. Calling it a coin flip is slightly unfair; calling it a reliable trading edge would be far more generous. The statistical significance is real, but the economic significance is small.

When strategies underperform the index

The backtest results are the most uncomfortable reading for anyone who believes news-based or correlation-based oil trading works. One strategy bought CVX based on the 20-day rolling correlation between its returns and Brent. It generated 77 closed trades with a 53% win rate and ended at +10.18% on $100,000. Meanwhile, SPY buy-and-hold did +68.30% over the same window — the strategy trailed by 58.13 points. A second strategy, triggered by daily counts of news articles mentioning Iran, Hormuz, or sanctions, did similarly: 75 trades, 53% wins, +12.01% final, trailing SPY by 56.30 points. Even the best single trade (+5.62%) and worst (-10.50%) on the news strategy show the asymmetry.

This pattern cuts against another favorite belief: that oil stocks should outperform when GDP growth accelerates. Over the past three years, in the five accelerating-growth quarters, XOM lagged SPY in four of them, averaging -12.4% excess. In the five decelerating quarters, it beat SPY in three, averaging +12.5% excess. The p-value is about 0.053 — suggestive but not definitive, and the direction is the opposite of the conventional cyclical logic.

The picture across these findings is consistent. Oil markets are full of plausible narratives — hedging, leading indicators, geopolitical news, macro cycles — that make for good conversation but poor prediction. The data leans toward a simple conclusion: the shortcuts are mostly noise, and the few real signals are too small to hang a strategy on. That might be the most useful finding of all.