Five Oil Backtests, One Lesson: High Win Rates Aren't the Same as Edge

A Saturday with nothing on the wire is its own kind of data point. The energy setups this platform tracks mostly trigger off news flow — geopolitical mentions, Brent momentum, sanctions chatter — and when the headline feed goes quiet, the triggers go quiet with it. Which makes this a good moment to look at what the week's research actually said, rather than what the tape was shouting.

The refining story got flipped

The most interesting result of the week runs against the intuition most energy watchers carry. The setup: a geopolitical oil headline spike with Brent already in positive momentum. The expectation was that refiners like Valero would lag the broad energy sector in that scenario — the familiar refining-compression story. Across 60 qualifying sessions, the opposite happened: VLO beat XLE by an average of +2.50% over the following 20 sessions, with a median of +3.12%, and VLO underperformed in only 25% of cases. That is a wide gap between what the prior says and what the sample says.

Worth being careful here. This is a conditional average over 60 sessions, not a law, and the finding was flagged as cutting against the hypothesis rather than confirming a durable rule. But when the direction comes out opposite the expected one, it should make anyone holding a confident crack-spread view pause.

Win rates mislead

The batch of backtests published through the week illustrates something that gets lost in strategy marketing. Take the BP setup: buy at the close when BP closes down at least 1.5% with Brent confirming. It returned +3.04% on $100,000 across 29 closed trades with a 78% win rate. Best single trade +3.69%, worst -15.18%. Nearly eight in ten trades right, and the whole exercise still produced three percent — trailing SPY buy-and-hold by 82.24 points over the same window. One outlier loss did the damage a high hit rate is supposed to protect against.

Now the mirror image. The BKR setup returned +29.56% across 66 trades with a win rate of just 48% — a coin flip — but a best trade of +16.81% against a worst of -5.32%. The shape of the distribution, not the hit rate, did the work.

USO is the busiest of the group: +74.14% across 112 closed trades, a 68% win rate, best +9.82%, worst -10.74%, finishing 11.13 points behind the benchmark. The XLE setup returned +8.71% over 60 trades at a 67% win rate, trailing by 76.57 points. The pattern across all of them: over these windows, none of the headline-triggered entries beat simply holding SPY. That is less a verdict on the signals than a reminder of what frequent in-and-out trading costs you against a benchmark that just sits there.

The edge that behaved was small

The XOM versus XLE study is the one that came closest to its hypothesis, and also the most candid about its own weakness. When Brent's 20-day return was negative and geopolitical mentions were elevated, XOM outran XLE over the following 20 sessions by an average of 0.68%, with a median of 0.92%, ahead in 75% of cases. That compares with a baseline of +0.17% for the same pairing. Directionally correct — but there were only eight independent episodes, and 0.68% over eight observations is a thin reed. The signal points the right way; it just does not carry enough weight to justify much confidence.

What ties the week together is that every one of these setups is news-shaped on the way in and slow on the way out — a headline triggers the entry, and 20 sessions decide the outcome. With the feed blank today, none of the triggers are firing, and that is not a neutral fact. It is the condition under which these strategies sit idle. The research keeps saying the same quiet thing: the triggers are frequent, the edges are narrow, and the benchmark is a harder opponent than the headline suggests.