WMB vs XLE 20-session returns when 10Y yield drops >20bps over 20 sessions and Brent is above its 50-day SMA
The thesis was clean: falling 10-year yields should hand midstream a duration bid, and firm crude should pull an upstream-heavy energy basket along with it. So WMB ought to beat XLE. Over the past three years, it did the opposite. Across 22 signal sessions where the 10-year dropped more than 20 bps over 20 sessions while Brent held above its 50-day SMA, WMB's average 20-session forward return was -0.56% versus +2.38% for XLE — a mean spread of nearly three points in XLE's favor, with WMB winning only 6 of 22 sessions. On ordinary days, by contrast, the spread averaged +1.15% in WMB's favor, so the setup didn't just fail to help; it flipped the baseline pattern.
The full breakdown — charts, distribution, and statistical tests — is below.
Over the past ~3 years, when the 10-year Treasury yield falls more than 20 bps over a rolling 20-session window while Brent crude closes above its 50-day simple moving average, does WMB outperform XLE over the next 20 trading sessions? I expect midstream to catch a duration bid from falling long-end yields plus firm crude's sector tailwind, so WMB beats an upstream-heavy energy basket.
How this was measured
Daily closes for WMB and XLE were built from each session's last minute bar, then the 10-year Treasury yield and Brent crude series were mapped onto that common trading calendar. A signal day requires the 10Y yield to be more than 20 basis points lower than it was 20 sessions earlier and, simultaneously, Brent crude to close above its 50-session simple moving average. The outcome is the close-to-close forward return for the following 20 sessions; WMB outperformance is measured as the arithmetic spread WMB minus XLE. Signal-day spreads are tested against zero with a one-sample t-test and against non-signal sessions with Welch's t-test.
The key numbers
Reading the numbers
Condition met only 22 of 683 sessions (3.2%). On signal days WMB's average 20-session return was -0.56% vs +2.38% for XLE, and WMB beat XLE just 27.3% of the time — so the expected duration-plus-crude tailwind did not show up.
The charts
On signal days, the WMB bar is below zero (-0.56%) while the XLE bar is positive (+2.38%), so the WMB-minus-XLE bar lands at -2.94% — the opposite of the hoped-for midstream edge. On non-signal days WMB averaged +2.65% versus XLE's +1.49%, and the spread flips to +1.15%. That contrast is the core finding: this exact yield-plus-crude condition was associated with WMB lagging XLE, not beating it.
The distribution of signal-day WMB-minus-XLE spreads is centered at -2.94%, and the left tail reaches -13.64% while the best outcome is +11.93%. With only 22 signals and a 27.3% win rate, the typical trade was WMB trailing XLE, not catching a duration bid. The histogram's negative center reinforces the bar chart: losers were more common than winners under this setup.
All signal days cluster in a narrow band of 10-year yield moves, between -0.33 and -0.20 percentage points, with an average of -0.2364 percentage points (about -23.6 bps). The corresponding forward WMB-minus-XLE spreads range widely, from -13.64% to +11.93%, but the mean stays at -2.94%. So even though every point meets the big-drop threshold, the forward spread did not systematically turn positive as the question hoped.
20-session forward return summary
| Group | N | WMB mean | XLE mean | WMB-XLE mean | WMB>XLE hit rate |
|---|---|---|---|---|---|
| Signal sessions | 22 | -0.0056 | 0.0238 | -0.0294 | 0.2727 |
| Non-signal sessions | 661 | 0.0265 | 0.0149 | 0.0115 | 0.6082 |
Signal sessions detail
| date | 10Y 20d change (pp) | Brent close | Brent 50d SMA | WMB 20d fwd | XLE 20d fwd | Spread |
|---|---|---|---|---|---|---|
| 2024-01-11 | -0.22 | 80.21 | 80.11 | -0.0228 | 0.0124 | -0.0352 |
| 2024-06-18 | -0.22 | 84.79 | 83.78 | 0.0447 | 0.06 | -0.0153 |
| 2024-06-24 | -0.22 | 86.75 | 83.46 | 0.0137 | -0.0107 | 0.0245 |
| 2024-06-25 | -0.23 | 86.81 | 83.34 | -0.0213 | -0.0109 | -0.0103 |
| 2024-06-26 | -0.22 | 85.76 | 83.23 | -0.0181 | 0.0173 | -0.0354 |
| 2024-06-27 | -0.32 | 87 | 83.15 | 0.0121 | 0.0151 | -0.003 |
| 2024-08-12 | -0.33 | 83.06 | 82.92 | 0.0159 | -0.0505 | 0.0663 |
| 2025-02-10 | -0.26 | 76.23 | 76.21 | -0.0245 | -0.0321 | 0.0076 |
| 2025-02-11 | -0.25 | 77.65 | 76.27 | 0.0146 | -0.035 | 0.0496 |
| 2025-06-20 | -0.2 | 78.73 | 67.15 | -0.0413 | -0.0373 | -0.004 |
| 2025-06-23 | -0.2 | 74.34 | 67.3 | -0.0375 | 0.0112 | -0.0487 |
| 2025-06-24 | -0.21 | 69.13 | 67.37 | -0.0561 | 0.0223 | -0.0784 |
| 2025-06-26 | -0.21 | 68.57 | 67.43 | -0.0792 | 0.014 | -0.0932 |
| 2025-07-01 | -0.2 | 67.63 | 67.46 | -0.0228 | 0.0234 | -0.0462 |
| 2025-09-16 | -0.3 | 69.69 | 68.97 | 0.0754 | -0.0297 | 0.1051 |
| 2025-09-17 | -0.24 | 69.19 | 68.91 | 0.0873 | -0.032 | 0.1193 |
| 2026-02-18 | -0.21 | 71.78 | 65.95 | 0.0153 | 0.0675 | -0.0522 |
| 2026-02-23 | -0.21 | 71.9 | 66.5 | 0.0069 | 0.0863 | -0.0794 |
| 2026-02-26 | -0.24 | 71.66 | 67.03 | 0.0027 | 0.1265 | -0.1238 |
| 2026-02-27 | -0.27 | 71.32 | 67.23 | -0.0063 | 0.1253 | -0.1316 |
| 2026-03-02 | -0.21 | 77.24 | 67.57 | -0.0406 | 0.0957 | -0.1364 |
| 2026-03-03 | -0.23 | 83.28 | 68.03 | -0.0406 | 0.0849 | -0.1254 |
The takeaway
The short answer is no: this setup has actually been a bad omen for WMB versus XLE over the past three years, not the tailwind the thesis expected. Across only 22 signal sessions, WMB's mean 20-session forward return was -0.56% versus +2.38% for XLE, leaving an average spread of -2.94% in XLE's favor; the median was -3.53%, and WMB won just 6 of those 22 sessions. On ordinary non-signal days the spread was positive, averaging +1.15% with WMB ahead about 61% of the time, so the condition flips the usual pattern. With 22 events this isn't conclusive, but it's a clear lean rather than a coin flip: there's roughly a 7-in-100 chance the negative average is pure luck, and the contrast with non-signal days would be a fluke only about 1.6 times in 100. So the practical takeaway is the opposite of the starting thesis — falling 10-year yields plus firm crude did not help WMB catch up to XLE; if anything, the signal leaned toward XLE. Treat this pattern with skepticism, and don't build a trade on it without a lot more evidence.
The fine print
- Only 22 triggers out of ~683 sessions; overlapping events (e.g., the June 2024 cluster) make the t-tests approximate.
- Macro data carry a T+1 publication lag; a tradable version would trigger a session after the observed close.
- Returns are close-to-close over 20 sessions, with no transaction costs or slippage; no threshold sensitivity checks were run.
- WMB vs XLE is a midstream-vs-upstream-heavy energy comparison, not a market-neutral or sector-agnostic edge.