Energy Signals Keep Winning Trades and Losing to the Benchmark

Five oil-linked studies crossed our desk this week, and they tell one story with almost suspicious consistency: the setups work often enough to feel good and still finish behind a plain index fund. With no headlines landing on the wire today, the research is the whole conversation, and the conversation is about the gap between winning trades and winning money.

The Win-Rate Trap

Start with the looker. Buying VLO at the close when Brent closes up more than 1.5% on the day returned +53.48% on $100,000 across 45 closed trades — a 76% win rate, which is the kind of number that gets a strategy screenshotted and shared. Over the same window, SPY buy-and-hold returned +76.34%. The strategy trailed by 22.85 points. Its best single trade was +6.14%; its worst was -7.86%.

That last pair is the whole story. A 76% win rate with a best trade barely bigger than its worst trade means the wins are numerous and small, and the occasional loss does real work. High hit rates flatter the psychology of a system without saying anything about its arithmetic.

The rest of the week's energy backtests repeat the pattern with different entries. Buying PSX when its 10-day total return runs more than 4 percentage points ahead returned +5.85% over 26 trades, a 54% win rate, trailing SPY by 70.48 points, with a best trade of +6.57% and a worst of -10.09%. Buying DVN on a close above its prior 20-day high while Brent cooperated lost money outright: -5.69%, 36% win rate, 22 trades, 82.03 points behind the benchmark — despite a best trade of +13.67% and a worst of -11.81%. Buying KMI when Brent fell more than 2% but KMI held up returned +22.64% across 50 trades at a 54% win rate, still 53.69 points behind, best +8.76%, worst -5.28%.

Four different conditions. Long into strength, long into a breakout, long into relative resilience. All four positive or near-flat in isolation, all four losing to the index by margins between 22 and 82 points. That looks less like four broken rules and more like one structural problem: a long-only, intermittently-invested single-name strategy has to clear a benchmark that compounded 76.34% over the sample. Most of them don't, and the ones that do need a payoff distribution far more skewed than a 76% hit rate suggests.

What the Beta Actually Confirms

The most useful finding this week wasn't a strategy at all. XOM's daily beta to Brent is 0.26 on down days and 0.16 on up days. That's the direction most people expect — oil equities take the escalator down and the stairs up — but the 0.10 gap isn't statistically clean, so it reads as a lean, not a proven asymmetry. Anyone treating it as a hedge is ahead of the evidence.

What is not in question is the linkage itself. The up-day beta of 0.16 carries a t-stat of 3.34 and a p-value of 0.0008. XOM's sensitivity to crude is real in both directions. The problem for signal-builders is that a real sensitivity and a tradable edge are different objects.

The One That Nearly Survives

The XLE test is worth flagging precisely because it almost clears the bar. On 55 qualifying days — lagged Brent up more than 2% with XLE down — the energy sector fund averaged +0.47% of excess return over SPY across the following 10 sessions, finishing ahead 60% of the time, against a non-signal baseline that is itself negative. The setup doesn't hold up as a standalone edge; the tilt is small enough to be indistinguishable from noise.

And yet it's the most interesting entry on the list. It's a relative bet rather than a directional one, it doesn't need energy to go up, and it doesn't need to outrun a 76.34% benchmark in absolute terms. If there's a durable edge in oil-linked equities, the week's research leans toward it living in the spread between two tickers rather than in a call on crude.

The takeaway across all five studies isn't that oil doesn't move energy stocks. The t-stat says it does. It's that the translation from crude to equity is noisy enough, and the benchmark high enough, that directional rules built on top of it keep producing satisfying trade logs and unsatisfying returns.