The trade that worked for the wrong reason

A position closed this week at roughly +1.4R. By the only measure most people publish, that is a win. By the measure that matters, it is a process failure that happened to pay, and those are more dangerous than ordinary losses because they teach you the wrong lesson.

Here is the review in the format I use for everything.

The thesis, as written at entry

Price two-plus standard deviations below a flat 50-day mean. Decline was a grind, not a gap. Estimates stable. Sector exposure was within cap. Planned risk: 0.75R on the first tranche, with a defined add level and a total planned risk of 1R across both. Exit conditions written before entry:

• Target: z-score returns to roughly zero.
• Invalidation: the 50-day mean itself falls more than a defined threshold.
• Time stop: roughly the estimated half-life of the deviation.

What actually happened

The position went almost nowhere for a week and a half, then gapped up on a day when the entire industry group gapped up on news about a different company in the same group. The z-score closed the gap in two sessions. I took the target, because that was the written rule and the rule fired.

Why it grades as a failure

The target condition triggered, so the exit was correct. The problem is upstream: the gap did not close because the thing I identified reverted. It closed because a sector-wide bid lifted everything in the group, including names that had never been stretched at all.

That distinction is invisible on the P&L and central to whether the method works. If I bank this as evidence that the setup works, I have added one observation to a sample that does not actually contain it. Do that a dozen times and you have a strategy you believe in for reasons that are not in the data.

The honest accounting: correct process, correct exit, no evidence gained. The trade goes in the log with the outcome tagged as exogenous, and it is excluded when I next measure the setup’s hit rate.

What I’d do differently

Nothing about the entry, the size, or the exit. All three followed the plan.

One thing about the review discipline: I nearly did not write this up, because it closed green and green trades do not itch. That is the bias in miniature. Winners get less scrutiny than losers, so the errors that hide inside winners survive longer.

The rule going forward: every closed position gets tagged at review time as thesis-driven or exogenous, and only thesis-driven outcomes count toward evaluating the setup. It costs nothing to record and it keeps the sample honest.

The general lesson

Expectancy math is only as good as the sample feeding it. A win rate computed over trades where half the outcomes had nothing to do with the signal is a number that describes the market’s mood, not your edge.

Nothing here is investment advice or a recommendation. It is my own process, published including the parts that did not work.