A Quiet Tape, an 80% Win Rate, and a Benchmark That Still Wins

Some days the news flow does the work. Today the headline feed came back empty, which is its own kind of datum: no catalyst, no narrative to ride, just the research stack and whatever the tape decides to do. So it's worth talking about what the models have been saying, because this week's batch is unusually coherent, and unusually uncomfortable for anyone who likes a tidy story.

Calm tape, better transmission

Start with the finding that cuts against instinct. The team asked whether XLE's one-day move carries more information about next-day Brent when energy is volatile. The front-running-in-stress thesis says yes: when things are wild, everything leads everything. The answer is no, and the tilt runs the other way. The lead-lag beta was 0.71 in the low-volatility regime versus 0.22 in the high-volatility one, a gap of roughly -0.50, the wrong sign for the thesis. The raw correlations tell the same story, at 0.25 in calm periods.

The read here is that stress doesn't sharpen signals, it smears them. In a calm regime, XLE's move is mostly about crude. In a violent one, it's also about positioning, flows, margin, and a dozen other things, so the crude content gets diluted. Anyone assuming volatility equals information density has it backwards, at least in this pair.

The 80% win rate that still trailed

Two energy backtests landed the same week, and both are lessons in what a hit rate actually buys you. The EOG range-break strategy returned +3.81% on $100,000 across 15 closed trades with an 80% win rate. Over the same window SPY buy-and-hold returned +85.28%, so the strategy finished 81.47 points behind. Best single trade +3.75%, worst -4.15%.

Eight out of ten winners, and it still doesn't matter. That isn't a bug in the backtest; it's the arithmetic of small edges, small trade counts, and a benchmark that spent the window compounding. The XOM rule, buying the close after a 1% down day with Brent as a filter, is worse in the other direction: -36.85% across 25 closed trades with an 8% win rate, trailing by 122.13 points, worst single trade -5.77%. A KMI 10-day moving-average entry fared no better: -1.28% over 66 trades, 37% win rate, 86.56 points behind.

Three setups, three different logics, one common denominator. Whether a rule wins 8% or 80% of the time says almost nothing about whether it beats doing nothing. The spread over the benchmark is the number that matters; the hit rate mostly isn't.

Where the regime does flip

Not everything came back negative. The Fed angle shows a genuine lean. On the 37 days when Fed/FOMC headlines spiked and Brent was already rising, XLE beat SPY over the next 20 sessions by an average of 3.49%, with the typical day even better at 8.11% and XLE winning 68% of the time. On the other 409 days, the average excess was 0.03%. The caveat sits in the research itself: this reads as a lean, not a proven edge.

The same restraint applies to the EOG-versus-XLE split. Across the 78 days since late 2023 when Brent's 20-day return was positive and SPY's was negative, EOG beat XLE by an average of 0.38 percentage points over the next 20 sessions, +0.94% versus +0.57%. That spread averaged -0.70% on all other days, so the regime flips the relationship by roughly 1.1 points. The sign is right; the magnitude is too thin to lean on.

Put it together and the message is consistent. The regime, not the setup, is doing the work. The calm tape is where the transmission lives, not the panicked one. The benchmark is the bar, not the win rate. And when the headline feed empties out, as it did today, the absence of a story is not the absence of a signal.