The Oil Signals Work. They Just Don't Beat Buy-and-Hold.

The research desk spent the week stress-testing energy signals, and the results have a consistent shape: almost every setup worked in the narrow sense, and almost none of them worked in the sense that matters. High hit rates, small payoffs, and a benchmark that keeps winning anyway.

Directional yes, edge no

Start with the cleanest-looking one. When Brent's 20-day return was negative and geopolitical mentions were elevated, XOM outran XLE over the following 20 sessions by an average of 0.68%, with a median of 0.92%, and XOM ahead in 75% of the cases. The baseline for that same pairing is +0.17%, so the setup more than triples it. That sounds like something. The catch is that it rests on eight independent episodes, and a 0.68% average monthly gap between a mega-cap integrated and the sector ETF is well inside the noise two large, correlated vehicles generate on their own. The finding's own wording is honest about it: directionally yes, but not by enough or often enough to call it a real edge.

The divergence setups fared worse. When Brent's 20-day return was negative while COP's was positive, COP actually lagged XOP over the next 20 sessions by an average of 0.41%, median shortfall 0.69%, coming out ahead in just 44.6% of the 121 trigger days, with a p-value of 0.32. The intuition that a single name running hot against a falling crude tape must eventually give it back does not survive contact with the data here.

The DVN setup is the most instructive of the three. After an up-close with a negative last 30 minutes while Brent was up half a percent, DVN trailed XOP by 0.67% on average over the next 10 sessions across 96 qualifying days, the laggard in 58% of cases. Then comes the deflation: DVN already lagged XOP by 0.20% on average across the broader baseline. The pattern isn't predicting anything. It's rediscovering a habit.

The benchmark is the argument

Three backtests landed the same week, and each one came with a scoreboard, which makes them far more interesting than the divergence work.

A rule that bought USO at the close when the 5-day average of Iran, Hormuz, sanctions and OPEC headline mentions was elevated returned +74.14% on $100,000 across 112 closed trades with a 68% win rate. Over the same window, SPY buy-and-hold returned +85.28%. A headline-driven oil trade that wins two out of three times still finished 11.13 points behind doing nothing. Its best single trade made +9.82%; its worst lost -10.74%, so one bad outcome roughly erases one good one.

The XLE rules were worse. Buying XLE when Brent's trailing 5-day return exceeded the ETF's returned +8.71% across 60 trades at a 67% win rate, trailing SPY by 76.57 points, with a best trade of +3.41% and a worst of -7.95%. Buying XLE on a cross above its 20-day simple moving average returned +29.20% across just 19 trades at a 58% win rate, trailing SPY by 56.08 points.

Notice the gradient. The more often the strategy trades, the better the win rate looks and the worse the compounding looks. High win rates sitting on top of small winners and larger losers is the signature of a system paying real costs to harvest a very small edge.

What the week actually says

The lazy read is that oil signals don't work. The more useful read is that several of them are real but tiny, on the order of a fraction of a percent over a 10- or 20-session window, which is a finding and a problem at the same time. Costs, slippage and timing have to come out of that fraction, and what's left is thin.

It also matters that these headline-count rules depend on a text stream that is lumpy by nature. Today's headline feed came up empty, which is its own small reminder that the conditions producing those 112 USO trades may not resemble the conditions the rule faces next.

If there's a hierarchy in this batch, it runs like this: single-name-versus-ETF divergences are mostly noise, with one of them failing outright and the other dissolving into a pre-existing tendency; event-driven oil trades carry a genuine signal but a benchmark problem; and the most robust statement of the week is a baseline, not a strategy. Some tickers simply behave a certain way relative to their sector, and most patterns claiming to predict them are just that behavior wearing a costume.