Energy's Headline Trade Is Backwards, and the Backtests Keep Agreeing

A tape with no headlines worth trading is a good day to read the lab notebooks. This week's published research on energy names has a running theme, and it is not a comfortable one for anyone who trades the story: direction has been the least informative variable in the complex.

Headlines are the worst input in energy

The cleanest example is the Hormuz work. After days when Hormuz headline intensity hit the top quintile — 48 such days over three years — VLO's average 10-session return differential versus XOM was about +2.3 percentage points, against +0.65 on ordinary days. VLO lagged XOM in only about one in four of those spike events, and the +1.65-point gap carries a p-value around 0.04. The reflexive read — escalation hurts the refiner, favors the integrated major — is simply not what the tape did. The charitable interpretation is that headline intensity measures attention, not cash flows. Both names are levered to the same crude, and a crack spread does not read the news. If anything, the crowded "obvious" side of that trade looks like the one that gets paid to be wrong.

The setups that did carry information

The findings that hold up share a structure: they condition on correlation and macro regime, not on price. When XLE is high-correlated to Brent and low-correlated to SPY, across 100 qualifying sessions, XLE beat SPY by an average of 3.2 percentage points over the next 20 sessions — XLE +3.8% against SPY +0.6% — versus a -0.5% lag across the 348 days that did not qualify. That is a 3.7-point swing, with XLE ahead 63% of the time. Nothing about that signal says "oil is going up." It says the energy sleeve has decoupled from the index, and that decoupling has been worth something.

The companion finding makes the same point with a different lens. Under front-end inversion — fed funds above the two-year yield — and Brent above its 50-day moving average, XOM has beaten XOP over the next 20 trading days: +1.90% versus +0.38%, a +1.52-point spread that was positive 75.6% of the time. The caveat matters more than the number: those 41 observations are really only 12 separate episodes. Twelve clusters is a thin book of evidence, and the honest way to read it is as a regime description rather than a repeatable edge. The spread is real; the sample is small.

Momentum keeps getting run over

Meanwhile, every simple energy-momentum variant in the stack has bled. Buying CVX at the close after Brent rises three straight sessions returned -12.73% on $100,000 across 31 closed trades, with a 61% win rate, a best trade of +3.39% and a worst of -9.41%. Read that pairing carefully: a majority of trades won and the strategy still lost, because the losses were roughly three times the size of the wins. Over the same window, SPY buy-and-hold returned +76.34%, leaving the strategy 89.07 points behind.

The FANG version, buying after its 10-day total return beats SPY by more than 3 points, is worse: -31.44% across 21 trades on a 29% win rate, best +8.72%, worst -12.90%, and 107.78 points behind SPY. Even the TTE 10-day moving-average crossover, which did make money at +7.44%, finished 68.90 points behind the benchmark over eight trades and a 38% win rate.

The through-line is hard to miss. Three momentum-flavored strategies, three losses relative to simply holding the index; two regime-conditioned setups that pointed somewhere useful, one of them supported by barely a dozen independent episodes. In energy right now, the data leans toward reading structure — correlation, inversion, positioning — and treating the news flow as the noise it has repeatedly turned out to be.