A Week of Oil Backtests That Win Trades and Lose to SPY

Four energy-sector studies crossed the platform this week, and the strangest thing about them is how consistent the punchline is. Not that the ideas fail — several of them win more often than they lose. It's that winning trades and beating the market turn out to be separate questions, and the oil patch is unusually good at answering the first one.

Win Rates Are Not Edge

OXY: buy at the close when the 5-day RSI closes below 30, with a Brent crude condition attached. Result: +4.14% on $100,000 across 8 closed trades, a 75% win rate, best trade +3.81%, worst -5.09%. SPY buy-and-hold returned +76.34% over the same window, so the strategy trailed the benchmark by 72.20 points.

CVX: buy when the 20-day average true range sits in the top quartile. Result: -3.85% on $100,000 across 15 closed trades, a 60% win rate, best trade +3.88%, worst -7.65%, trailing the benchmark by 80.19 points.

MPC: buy when Brent crude closes down more than 2% on the day. Result: +11.36% across 37 closed trades, a 43% win rate, best trade +11.17%, worst -9.43%, trailing by 64.98 points.

Read those three in a row and the ordering inverts. The highest win rate produced the smallest gain; the lowest win rate produced the largest. Eight trades is not a sample, it is an anecdote with a spreadsheet, and fifteen is not much better. In the CVX case the worst trade was nearly twice the size of the best, which is the part that matters: these setups are being judged on hit rate when the thing that actually moves an equity curve is the shape of the losses.

Headlines Do Not Rotate Cleanly Into Tickers

The week's other two studies tested a more intuitive idea — that a loud upstream headline sends money rotating into one oil-adjacent name over another.

BKR versus OIH after OPEC, quota, and output-cut headline spikes. Across the 12 spike days that cleared the top-quintile bar, BKR averaged +0.97% over the following 10 sessions while OIH averaged +1.40%, a spread of roughly -0.43% and a median closer to -0.55%. The platform's answer was a flat no, and if anything the relationship ran the other way.

KMI versus XLE after Iran, Hormuz, tanker, and sanctions spikes. Across 17 such sessions over three years, KMI averaged +0.51% over the following five sessions against +0.23% on ordinary days — the opposite sign of the hypothesis, though the median of -0.06% is essentially dead flat, meaning the positive average rests on a couple of favorable instances. Also a no.

Headline intensity is seductive because it is loud, dated, and easy to count. But nothing routes an OPEC quota story into one midstream ticker and not its neighbor, and the data keeps declining to pretend otherwise.

The One That Actually Cuts

The keeper of the week is the USO versus XLE tail study, because it answers a question with a better question. On the 363 sessions where Brent's 5-day return was negative, USO's worst 5% of days averaged -4.19% against -2.62% for XLE. Its worst 1% of days ran -6.44% versus -3.92%. On the average downswing day, USO bled -0.46% against -0.19%.

The skewness framing, it turns out, was the wrong lens. These two are not shaped differently so much as scaled differently once you get deep into the tail — roughly half again as severe at the extremes. That is a more useful fact than a skewness coefficient, because it describes how far the floor drops rather than which way the distribution leans.

Taken together, the week reads as a caution about confusing narrative with mechanism. Energy has the best stories in the market and some of the least cooperative statistics. The studies that held up were the ones measuring how bad the bad days get, not the ones promising that the right headline would show up as the right ticker.