Crude's Direction Is a Weak Proxy for Energy Stocks, the Data Keeps Saying

Two energy backtests landed on the platform this week, and both started from the same premise — crude goes up, buy an oil stock — and both ended up losing money while a passive benchmark compounded. That is the most useful thing published in the past few days, and it has nothing to do with where crude is heading next.

The first test buys SLB at the close whenever Brent settles above its 50-day simple moving average. Across 31 closed trades it returned -54.21% on $100,000 of starting capital, with a 26% win rate. SPY buy-and-hold returned +76.34% over the same window, so the strategy trailed the benchmark by 130.55 points. Best single trade +4.66%, worst -12.74%.

The second buys CVX after Brent closes higher three sessions in a row. Better hit rate, 61%, and still a loss: -12.73% on the same starting capital, trailing SPY by 89.07 points. Best trade +3.39%, worst -9.41%.

A 61% Win Rate That Still Lost Money

Read those side by side and the lesson isn't that oil equities are uninvestable. It's that a commodity's direction is being used as a proxy for an equity's behavior, and the proxy is weak. The CVX case is the sharper one precisely because the win rate was respectable: 61% of trades made money and the strategy still finished down double digits. Wins too small, losses too large. A signal that produces that shape is not really a signal — it's a coin flip with an unattractive payoff attached.

Note also what the trigger was: a filter on the commodity, not on the equity. Brent above its 50-day average says something about crude. It says almost nothing about how an oil-services stock will behave over the following weeks, and the two backtests are the receipt.

The Interesting Version of the Question

Where this week's research actually leans is not in crude's direction but in the conditions surrounding it. Three findings point the same way.

XLE versus SPY, when XLE is highly correlated to Brent and only weakly correlated to the S&P: across 100 qualifying sessions, XLE averaged +3.8% over the next 20 sessions against +0.6% for SPY — a 3.2-point spread, with XLE ahead 63% of the time. On the 348 days that didn't qualify, XLE lagged by -0.5%. That's a 3.7-point swing in the expected direction.

XLE versus SPY after volume-confirmed down-crude capitulation days: across 71 such days, XLE beat SPY by +0.88% over the following 10 sessions. Remove the volume confirmation and the same setup gives back -0.42%; all down-crude days average -0.17%. The volume filter appears to be separating something from ordinary weakness rather than manufacturing it.

And in relative-value terms: when Brent dumps 5% while OIH is still up on a five-day view, OIH averaged +4.22% over the next 20 sessions versus +2.11% for XOP across 16 qualifying days since September 2023. Separately, EQT — when its 20-day Brent correlation was negative while its own 20-day return was positive — beat XLE by +2.16% over the next 20 days, against -0.44% on all other days. That's a +2.60-point edge with a day-level p of 0.010.

Lean, Not Proof

The caveats matter, and the research states them. Sixteen qualifying days for the OIH/XOP setup is a thin sample, and the finding itself declines to call it a real edge rather than luck. The EQT p-value looks strong at face value, but the win rate was only modest, which is exactly the shape that makes a single significant test less impressive than it reads. And a p of 0.010 found among a batch of published tests is not the same animal as a p of 0.010 tested once and once only. The XLE-versus-SPY regime finding has the deepest sample of the group, and even there the language is "lean," not "edge."

The through-line is consistent, though. The market doesn't seem to pay much for crude's direction, and it does seem to pay — moderately, inconsistently — for context. Whether energy is trading on its own story or on the macro one. Whether a down day arrived with volume behind it. Whether the relative move inside the energy complex is diverging from the commodity itself. Those are second-order questions, and second-order questions are where the data stops shrugging.

The two backtests are the reminder that the first-order question — is Brent up or down — has been asked repeatedly, and for these particular strategies the answer was a loss of more than half the capital in one case and a near-miss in the other. The conditions are where the lean lives. Not an edge yet. Just a better question.