EQT vs XLE forward 20-session returns after extreme Brent trailing-20d regimes
A 2.5-point relative move is the kind of gap that gets your attention. After Brent's trailing 20-day return lands in the top quintile — a gain of roughly 6.3% or more — EQT averaged -0.58% over the following 20 sessions while XLE averaged +1.90%. That's the decoupling the thesis predicted: oil shocks lift oil-beta energy equities, not the gas producer.
Then the sample hollows it out. Those 133 event days are overlapping windows, and once they collapse to 15 genuinely independent events, the paired t-stat is -0.77 and p = 0.45. Worse for the story, EQT also trailed XLE by 0.64% after crude crashes, so there's no downside cushion either.
The full regime breakdown, and the test comparing the two, is below.
Over the past ~3 years, when Brent crude's trailing 20-day return is in the top quintile of its distribution, does EQT's next 20-session return lag XLE? I expect natural-gas producers to decouple from oil supply shocks, so EQT should trail oil-beta energy equities after crude spikes while holding up better when crude crashes.
How this was measured
Daily closes for EQT and XLE were built from minute bars; Brent's daily series was aligned to stock trading days by forward-fill. The Brent trailing-20-session return was computed on that aligned series. Days were split into top quintile, middle 60%, and bottom quintile using the full available forward-eligible sample. For each regime, the next 20-session return for EQT and XLE was recorded and the signed EQT-minus-XLE gap was summarized. Because overlapping 20-session forward windows inflate independent observations, paired inference was also run on non-overlapping events separated by at least 20 sessions.
The key numbers
Reading the numbers
Brent's 20-day return topped 6.27% on 133 sessions, and over the following 20 sessions EQT averaged -0.58% while XLE averaged +1.90% — a 2.47-point shortfall. But once overlapping windows are stripped out only 15 independent episodes remain, and p=0.45 means that gap is well within what random noise could produce.
The charts
This groups the 665 days by where Brent's trailing 20-day return sat and shows what EQT and XLE did over the next 20 sessions. The bar to fix on is EQT's rightmost one: it is the only negative bar on the chart, at roughly -0.6%, while XLE's top-quintile bar sits near +1.9%. Note the reversal in the middle 60% bucket, where EQT (+1.9%) actually beats XLE (+1.3%), and the bottom-quintile pair is close (EQT +1.2% vs XLE +1.9%). So the decoupling you expected shows up directionally only after crude spikes, and it is mild rather than dramatic.
This plots the EQT-minus-XLE spread itself, so anything below the zero line means EQT trailed. The top-quintile box is the one that matters: it sits below zero, with the middle of the distribution running roughly -9% to -7% and a mean of -2.47%, while the middle-60% group has a positive mean (+0.65%) and the bottom-quintile group is barely negative (-0.64%). The tell is the spread of the whiskers — top-quintile outcomes run from about -15% to +17% — which is why an average gap this size fails to clear statistical significance despite pointing the way you predicted.
Forward-20-session return summary by Brent trailing-20d regime
| Regime | Raw N | Clean N | EQT mean | XLE mean | Gap (EQT-XLE) | Clean gap p-value |
|---|---|---|---|---|---|---|
| Bottom quintile | 133 | 15 | 0.0124 | 0.0189 | -0.0064 | 0.9057 |
| Middle 60% | 399 | 30 | 0.019 | 0.0125 | 0.0065 | 0.7097 |
| Top quintile | 133 | 15 | -0.0058 | 0.019 | -0.0247 | 0.4519 |
The takeaway
Your direction is right but the evidence is too thin to act on: after Brent's trailing 20-day return sits in the top quintile (a gain of roughly 6.3% or more), EQT averaged -0.58% over the following 20 sessions versus +1.90% for XLE — a 2.5-point lag for the gas producer, exactly the decoupling you'd expect. Strip out the overlapping forward windows, though, and that gap rests on just 15 genuinely independent events; p = 0.45 means it's a coin flip, not a signal. The other half of the thesis doesn't survive at all — after crude crashes, EQT also trailed XLE, by 0.64% (p = 0.91), so there's no 'holds up better on the downside' behavior to speak of. And the two regimes aren't different from each other (p = 0.69), which is the cleanest way of saying the whole pattern is noise. The 2.5-point relative move is big enough to be interesting, but with 133 event days collapsing to 15 clean observations over a three-year window, you cannot tell it apart from luck. Takeaway: treat this as an untested hypothesis rather than a result — the sample can't confirm or refute the decoupling story, and nothing here justifies leaning on crude regime as a reason to favor XLE over EQT or vice versa.
The fine print
- Only 15 non-overlapping events per regime — thin evidence, and the overlapping 133-day counts massively overstate how much independent information is here.
- Quintile cutoffs are set on the full sample, not point-in-time, so the regime labels are in-sample and mildly self-selecting.
- Brent is forward-filled onto US trading days, which can shift a crude move into the next EQT/XLE session and slightly misalign the 20-session trailing window.
- EQT is a gas name and XLE is broad energy — weather, storage, and basis differentials can easily dominate a 20-session horizon regardless of crude.