Five Energy Backtests, Two Winning Records, and Not One Beat the Index

The news wire is quiet today, which makes it a good moment to look at what the platform's own energy backtests have been saying for the past three sessions. The verdict is not subtle. Five setups across oil services, midstream, refiners, gas and crude-linked names, and the best of them still finished 67 points behind a passive index.

A win rate is not an edge

Start with the one that reads best on a dashboard. Buying OIH at the close after it closes down at least 1.5% from the prior close returned +8.86% on $100,000 across 47 closed trades, with a 57% win rate. That looks like a functioning strategy until you put it next to SPY buy-and-hold over the same window, which returned +76.34%. The dip-buy trailed the benchmark by 67.48 points, with a best trade of +8.10% and a worst of -6.11%.

The same arithmetic shows up in COP, where buying weakness in the bottom 20% of its prior 20-session range produced a +5.57% return and an 80% win rate over 10 closed trades. Eight out of ten right, and still 70.77 points behind. Ten trades is a small sample, and an 80% hit rate on ten trades is two losses away from looking ordinary. But the direction of the gap is the point: the trades were frequently right and the result was still wrong relative to doing nothing.

The setups that never found the upside

EQT is the cautionary tale of the batch. Buying after Brent crude closes up more than 1.5% on the day produced 63 closed trades, a 0% win rate, and -74.09% on $100,000. The best single trade was -0.03%. The worst was -10.94%. It trailed the benchmark by 150.43 points. Zero for sixty-three is not variance; it is a trigger that either points the wrong way or gets consumed by the cost of acting on it that often.

The smaller samples point the same direction. WMB, bought when its daily range was the narrowest of the prior five sessions, went 0-for-5 for -9.67%, with a best trade of -0.26%. The XLE 10-day moving-average crossover went 20% wins across 10 trades for -10.86%, best trade +3.07%, worst -3.29%. Neither sample is large enough to call proof of anything — but neither ever found a clean winner either.

The near-miss worth watching

The most interesting result is the one that did not clear the bar. Across 33 refined-margin momentum events, MPC outperformed XLE over the following 20 sessions in 54.5% of cases, with a mean excess return of +0.49% (MPC +3.17% versus XLE +2.68%). The one-sample p-value came in at 0.7030, nowhere near the 5% threshold. Worse for the story, relative to all eligible days the event edge was -1.22%, meaning the conditional days were actually worse than a randomly chosen day.

That is the pattern to internalise from this batch. A 54.5% hit rate and a positive mean look like something. Statistically they are a coin flip with a favourable rounding error, and the overlap in forward windows flatters the sample size. The magnitude is meaningful only if you accept a sample that does not support it.

What the pattern says

Energy is a sector where mechanical entry triggers keep looking productive in isolation and then lose to an index that simply sat there. Two of these five strategies made money, and both trailed by more than 67 points. Three lost money outright, one of them catastrophically. The common thread is not the direction of the signal — it is that the trigger, on its own, carries almost none of the information the returns require.

The data leans toward a blunt reading: in this complex, the useful variable is not when to enter. It is how the position is sized, when it is exited, and what regime it is entered in. The EQT result, with 63 attempts and not one winner, is what happens when a signal gets repeated into a market that is not paying for it.

For now the news flow offers nothing to override that. The backtests are the story, and the story is that energy keeps handing out win rates that do not convert.