Crude Signals: One Rare Winner, Many False Leads

Every week, the research desk at trades.run throws another crude-oil-linked strategy at the historical tape, and every week the tape answers with a shrug. The latest batch of backtests is no exception: one clear outperformer, a handful of modest edges, and a couple of outright traps. The message, once you read across the findings, is that crude-market timing is less about the direction of Brent and more about the specific instrument, the fundamental context, and a healthy dose of skepticism.

A Rare Outperformer

The standout comes from a strategy that buys KMI at the close after Brent falls more than 1% in a day, with an additional filter that the finding doesn't fully spell out. Whatever the extra condition is, it worked: across 49 closed trades, the backtest turned $100,000 into $102.95% total return, with a 61% win rate. Over the same window, SPY buy-and-hold returned +68.30%, so the strategy beat the benchmark by 34.65 percentage points. Its worst single trade was -5.16%, and its best was +11.01%. That is the kind of edge quants dream about — but notice that it is attached to one midstream name, not the whole sector.

Contrast that with two sector-level trend-following backtests published the same week. Buying XLE when Brent closes above its 100-day moving average produced a +7.29% return over 28 trades, with a 46% win rate — trailing SPY by 61.01 points. Buying MPC when Brent closes below its 20-day average fared only slightly better: +9.26% over 38 trades, but with a dismal 37% win rate and a -59.05 spread versus the benchmark. The pattern is consistent: crude-level trend gates, whether above or below, delivered weak or negative alpha when applied to the broad sector or a single refiner.

The Seasonal Edge That Flipped

The research also pokes a hole in a pet theory about refiners and summer. A test of VLO's May-September forward 10-day return versus XLE found that when Brent was above its 50-day moving average, the summer edge was actually negative: -0.91% versus +1.74% in winter, a swing of -2.65 percentage points. When Brent was below trend, the summer edge turned positive but modest at +0.33 points. So the conventional narrative — strong crude means strong refiner seasonality — gets no support here; if anything, it is backwards.

Another finding on PSX looked at news sentiment on days when Brent fell more than 1% versus rising-crude days. Top-quintile sentiment on down-crude days averaged +2.21% over the next five trading days versus +1.63% after rising-crude days, a +0.58 percentage-point gap. But with only 18 down-crude days in the sample versus 36 rising-crude days, the statistics are far from conclusive. The direction is suggestive, but the data leans toward "maybe," not "yes."

The Modest Signal That Holds

Perhaps the most useful finding is the one with the least dramatic headline. ET, the midstream operator, has tended to outperform XLE in the 30 days after a quarter-over-quarter free-cash-flow margin expansion, but only when Brent sits below its 50-day moving average. The edge is real but small: on the 76 signal days, ET beat XLE by an average of 3.35 percentage points, versus 0.99 on non-signal days. The median excess return was just 0.74 points. It is not a home-run strategy, but it is consistent, and it survives a condition—the crude trend—that many traders would expect to dominate.

Taken together, this run of research tells a coherent story. The one big winner is an idiosyncratic, event-driven buy on a single stock after a sharp crude drop. The trend-following sector plays failed. The seasonal and sentiment edges are weak or reversed. And the fundamental margin signal, while modest, is the only one that holds up across both a directional and a median basis. The data lean toward this conclusion: in energy quant trading, the edge is in the specifics, not in the crude tape. Beta is easy to manufacture; alpha is not.

None of this is investment advice. It is simply what the historical backtests say — and even those, as the PSX result reminds us, can be running on a sample of eighteen days. The next time a crude-driven signal looks obvious, the smartest move might be to ask what the tape actually did, not what the narrative says it should have done.