Energy's Best-Looking Signals Keep Failing Their Own Tests

The research desk ran a batch of energy setups this week, and the results share a shape: the intuitive version of each trade didn't just fail to clear the bar, it frequently pointed the other way. That's a more interesting problem than a weak signal. A weak signal can be tuned. A signal that inverts is a story the tape doesn't believe.

Capitulation that looks like an ordinary session

Start with the loudest one. The idea was that top-quintile opening volume in XLE on days when Brent fell at least 1% marks a capitulation, so the next 20 sessions should favor XLE over SPY. Over the roughly two years the panel covers, 55 days met both conditions. XLE averaged +3.27%, SPY +2.69%. The spread, +0.58%, is statistically indistinguishable from zero at p ≈ 0.60. The 441 ordinary days logged +0.37%, p ≈ 0.85. That's the whole finding: panic volume on a down-crude day tells you about as much as any other session.

Two signals that lean the wrong way

XLE's volatility-spread compression is the more awkward one, because the setup inverts. On the 17 days when XLE's 20-day realized volatility sat in the bottom quintile of its spread to SPY and Brent's 20-day return was positive — a combination that reads like calm accumulation — XLE went on to lag SPY by an average of 3.39% over the next 20 sessions, median -1.26%, ahead in only 29% of cases. Roughly one in three.

The yield story gets similar treatment. Across the 40 sessions where the 2-year yield rose at least 30 basis points over 20 days while Brent's 20-day return was positive, the expectation was that an integrated major like XOM would hold up better than a shale name like FANG. It didn't. FANG beat XOM by 1.36% over the following 20 sessions. The honest reading is that both were weak — FANG roughly flat at -0.04%, XOM down 1.39% — so this looks less like a shale win than an integrated loss.

Averages and win rates that flatter

APA's positive earnings surprises during a falling Brent tape look compelling at first: seven qualifying events, APA beating XOP by an average of 4.82% (APA +5.98% against XOP +1.15%). Then you notice the median gap is 0.06% and APA outran the basket in only four of seven events. The average is two events wearing a trench coat.

The two backtests in the batch are a blunter version of the same lesson. The BP rule — enter at the close after a 1% down day — returned +1.25% on $100,000 across five closed trades with an 80% win rate. The VLO ratio setup, entering on a 20-day low in VLO versus USO, returned +16.75% on the same capital across eight trades, 75% win rate. Over the same window, SPY buy-and-hold returned +85.28%, leaving those strategies trailing by 84.03 and 68.53 points respectively. Five and eight trades is not a sample; it's an anecdote with a win rate attached.

The news tape isn't arguing the other side

The week's headlines offer little to push against any of this. Valero traded after-hours at EUR 342.00, up 0.04% from the prior close. TETRA Technologies set its Q3 report for November 2 with a call the next day. Eni is offering a 30% discount on electricity and gas rates through its retail arm. There's a Form 4 filing, a Tesla-crafted auto-vote plan the SEC cleared for wide use, and a 52-week low at $8.93 in a cable name down 23.72% over the past year. None of it touches crude, refining margins, or the yield curve.

What the batch suggests is narrower and more useful than "energy is untradeable." It's that this sector's narratives are being asked to carry more weight than the measured dispersions can bear. A 0.58% spread at p ≈ 0.60 and a 0.06% median don't mean nothing happens around energy. They mean what happens isn't the thing the story predicts.