Five Backtests Beat the Coin Flip This Week. All Five Still Trailed SPY.

A week with no headlines worth trading is a useful week to look at what the models actually did instead. The platform's published research this week produced four single-name setups with win rates between 50% and 73%, and one geopolitical hypothesis that fell apart under inspection. The results are worth reading together, because collectively they say something less flattering than any one of them says alone.

Five wins, one benchmark

Start with the scoreboard. The XOM-versus-CVX relative-strength setup returned +30.26% on $100,000 across 11 closed trades, winning 73% of them. The VLO volatility-compression setup returned +20.26% across 7 trades at a 57% win rate. The HAL breakout filtered on Brent added +16.92% across 6 trades. The FANG mean-reversion setup, buying the bottom quintile of 20-day returns, managed +3.32% across 6 trades with a 67% win rate.

Every one of those is a positive result. Every one of them also trailed SPY buy-and-hold over the same window, which returned +76.34%. The gaps are not small: 46.08 points for XOM, 56.07 for VLO, 59.42 for HAL, 73.02 for FANG.

That is the tension worth sitting with. These are not broken strategies — a 73% win rate with a best trade of +8.63% and a worst of -7.35% is a coherent distribution, and HAL's 50% win rate paired with a +12.42% best and a -2.33% worst is a textbook payoff asymmetry. What they have in common is that they are episodic. They fire, they sit in cash, they fire again. The benchmark does not sit in cash. Over a stretch where the index compounds relentlessly, intermittent edges get punished for their downtime, not their accuracy.

FANG is the outlier worth flagging for a different reason. A 67% win rate that produces +3.32% means the wins are the small ones — best trade +2.36%, worst -4.15%. Hitting often and still finishing roughly flat is a specific failure mode, and it looks different from HAL's occasional-big-winner profile even though both land in the same trailing-the-index bucket.

The Hormuz premium that fades

The week's cleanest negative result is the energy-and-headlines study. Across 36 days where Iran, Hormuz, tanker and sanctions coverage spiked into the top quintile, XLE beat SPY by an average of just +0.33% — and won only 56% of the time. On a 10-session horizon the average excess return flips to -0.45%.

The reflex trade here is well known: geopolitical friction in the Gulf, buy energy. The data says the day-one reaction is statistically indistinguishable from a coin flip and the follow-through is mildly negative. Nothing in that pattern supports treating a headline spike as a signal rather than a mood.

That is a fitting result for a week with no headlines at all. The tradeable content of a headline is rarely the headline itself.

Where the lean actually is

The one place a real asymmetry shows up this week is the Brent volatility study. On high-volatility days, LNG averaged +3.20% over the following 20 sessions versus +1.64% for XLE — a +1.56% spread in LNG's favor. LNG also won 56.6% of those event days against just 36.7% on non-event days, and the raw event-versus-non-event contrast of +1.92% carried a p-value of 0.009.

The platform's own conclusion was careful, and it should be: this is a lean, not a proven edge. One contrast, one p-value, a 20-session horizon. It is a hypothesis with a number attached, which is more than most rotation stories have.

Put the whole week together and the message is not that the models failed. It is that the bar is set by a benchmark that never stops compounding, and clearing a coin flip is nowhere near enough to clear it.