COP vs XOP forward 20 sessions after Brent-negative / COP-positive divergence
The setup sounds reasonable: if COP is grinding higher while Brent's 20-day return sits in the red, something company-specific is probably working — hedges, execution, a book that isn't pure crude beta. Over the past three years that read didn't pay.
Across 121 days where Brent was down on a 20-session basis and COP was up, COP went on to lag XOP over the following 20 sessions by an average of 0.41%, finishing ahead in just 44.6% of cases. The gap sits well inside the noise, and the year-by-year splits wander in both directions.
The full trigger construction, the comparison against falling-Brent days where COP wasn't up, and the rest of the evidence are in the analysis below.
Over the past ~3 years, when Brent crude's trailing 20-day return is negative but COP's trailing 20-day return is positive, does COP outperform XOP over the next 20 trading sessions? I expect COP's positive tape against falling crude to signal company-specific hedge books or operational execution that the oil-beta selloff is underrepricing, so it should lead high-beta E&Ps until crude stabilizes.
How this was measured
Daily closes for COP and XOP were resampled from minute bars and joined on common trading days. Brent daily was aligned to those trading sessions and lagged one session to respect the T+1 daily macro release convention; the Brent trigger is a 20-session pct_change on the lagged series, while COP 20-day return uses same-day close. The event is Brent 20d < 0 and COP 20d > 0. Forward outcome is close-to-close 20-session return for each name, and outperformance is COP minus XOP. We test the event sample against zero with a one-sample t-test and against falling-Brent days where COP was not up with a Welch two-sample t-test; windows overlap, so inference is descriptive.
The key numbers
Reading the numbers
Out of 669 trading days, 121 (about 18%) met the trigger of Brent down and COP up over the prior 20 sessions. On those days COP actually trailed XOP by 0.41% on average over the next 20 sessions and was ahead only 44.6% of the time, with a p-value of 0.32 — in plain terms, that result is close enough to a coin flip that it can't be distinguished from no effect at all.
The charts
This histogram stacks up all 121 trigger days by how much COP beat or trailed XOP over the following 20 sessions. The bulk of the bars sit just to the left of zero, and the average outcome is -0.41%, meaning the typical trigger day was followed by mild underperformance rather than the leadership the thesis expected. The tails are worth a look: losses stretch to -8.8% while gains reach +12.8%, so a handful of big winners pull the average up relative to the median of -0.69%. That gap between mean and median says the story is not 'COP usually wins slightly' but 'a few large wins sit on top of a majority of small losses.'
This side-by-side box plot asks whether the trigger day is actually special by comparing it with two benchmarks: falling-Brent days where COP was not up (250 days) and days when Brent was not falling (298 days). The event group averages -0.41%, the falling-Brent control -0.74%, and non-falling Brent roughly flat at -0.07%, so the trigger beats its control by only about 0.32 percentage points — a gap that overlaps heavily from one box to the next and carries a Welch p-value of 0.53. Notice too that every group's center sits at or below zero; the control group's left whisker even reaches -13.8%. The takeaway for the question is that a positive COP tape against falling crude did not reliably tr
Event sample by calendar year
| year | n | mean_cop_fwd20 | mean_xop_fwd20 | mean_outperf | median_outperf | frac_pos |
|---|---|---|---|---|---|---|
| 2023 | 12 | -0.0516 | -0.0529 | 0.0012 | -0.0054 | 0.4167 |
| 2024 | 31 | -0.0266 | -0.0111 | -0.0155 | -0.0211 | 0.3226 |
| 2025 | 65 | -0.0019 | -0.0059 | 0.004 | 0.0042 | 0.5385 |
| 2026 | 13 | -0.0081 | 0.0146 | -0.0226 | -0.0318 | 0.3077 |
| All | 121 | -0.0138 | -0.0097 | -0.0041 | -0.0069 | 0.4463 |
The takeaway
No — the setup didn't deliver. When Brent's 20-day return was negative while COP's was positive, COP actually lagged XOP over the following 20 sessions by an average of 0.41%, with a median shortfall of 0.69%. It only came out ahead in 44.6% of the 121 trigger days, so the typical outcome was COP trailing, not leading. The gap isn't statistically distinguishable from zero (p = 0.32), and comparing against falling-Brent days where COP wasn't up barely moves the needle — a 0.32% edge with a p of 0.53, which is essentially noise. Year by year it whipsaws too: COP trailed by 1.5% on average in 2024 and 2.3% in 2026, while 2025 flipped to a modest 0.4% lead, so there's no consistent company-specific signal here. The practical read is that COP's relative strength against a falling crude tape has not been a reliable tell that it will outrun high-beta E&Ps over the next month — if anything the tilt is mildly the other way, and this is close to a coin flip rather than an edge worth acting on.
The fine print
- 121 trigger days is a decent count, but the forward windows overlap heavily, so the effective number of independent bets is far smaller and the p-values flatter the certainty.
- These are price returns only — XOP pays out more in distributions than COP, so on a total-return basis COP's shortfall would look slightly worse, not better.
- The trigger day count swings a lot by year (12 in 2023, 65 in 2025), so a few clustered episodes are driving much of the average.
- 20-day lookback and 20-day forward horizon is one choice; 10-, 40- or 60-session versions could tell a different story.