Your decision was right. The outcome wasn't.

I read Annie Duke’s Thinking in Bets over the summer, and one habit from her professional poker years has stuck with me since: Annie went looking for the mistakes in her decision even on the hands she’d won.

Re-examining a win that closely is harder than it sounds. Winning feels like proof you did it right. Duke’s argument is that it isn’t proof of anything, because a decision and its outcome are two different things that just happen to run into each other. She has a word for the habit of grading the decision on the result instead of on the quality of the decision: resulting.

Winning hands with mistakes in them

Resulting is judging a decision by how it turned out instead of how it was made. Win the hand, and whatever you did on the way there quietly gets marked correct, however reckless your decisions might have been. Lose the hand, and a perfectly sound call gets rewritten as a mistake, because the only evidence anyone weighs afterward is the result.

A trading rule can fire exactly as written and still lose money on a given trade. That’s not the rule failing, it’s a good process meeting an unlucky draw from the market. Resulting is what happens when you can’t tell those two situations apart afterward, and start rewriting the rule based on an outcome your decision might not have been responsible for at all.

The branches that don’t survive

The part of the book that gave resulting extra weight for me wasn’t about poker at all, it was about decision trees. At the moment you decide, several futures are still open, each with its own odds. You choose a branch, and time moves forward. But once that branch resolves into what actually happened, the others don’t just lose, they vanish. Who keeps a detailed logbook of what would have happened if you’d chosen a different branch, or what would have happened if yet another branch had become reality? The tree collapses down to the single path that got walked, and that path is the only one your memory bothers to file.

Losing the branches is a sharper problem than ordinary hindsight bias. It’s not only that you misremember how confident you were. The actual context you’d need to judge the decision fairly, the odds you were weighing, the information you didn’t have yet, disappears along with the branches that didn’t happen. You’re not just biased looking back. You’ve lost the context you were in when you actually made the decision.

What Evaluate keeps that memory doesn’t

The tree metaphor is where the book stopped being about poker and started being about the Calibration card in Aurono’s Evaluate tab.

Every signal your strategy sees gets a reason attached the moment it happens, taken or not. Not “2 bought, 0 sold,” but the sell line crossed sixteen times, and eleven of those stayed a non-event because the price sat below cost basis, and five because there was nothing left to sell. Those are branches that didn’t happen, and Aurono writes them down anyway, before anyone knows whether not-selling was going to look smart in hindsight or not.

Logging all these events is resulting run in reverse. The record isn’t reconstructed from how things turned out, it’s captured at decision time and left alone. In the ONDO case study I wrote earlier this year, three sell signals fired below cost basis and got blocked. The line that made the post worth writing was “the signal was real, the guard held,” and it only means anything because that block was logged the moment it happened, not recovered later from a chart that happened to end up green. In other words: the strategy is doing well, it’s sitting on a profit, but that still leaves room to go looking for mistakes in it.

Review the wins too

Duke’s habit of re-checking hands she’d won is worth stealing on its own, no tooling required. Next time a trade closes green, ask the trade the same question you’d ask a loss: were your rules actually good rules, or did the market bail out a flawed decision, and were you basically just lucky? A strategy that sells early into a lucky spike isn’t vindicated by the spike. The Audit trail keeps the same trade-by-trade record either way, so that question has an answer instead of a guess wearing memory’s clothes.

The quality of a decision was never something the outcome could tell you. The quality of a decision was set the moment you wrote the rule, back when the branches of every alternative were still open and you didn’t yet know which one would become reality. Keeping that moment on the record, instead of letting the outcome quietly rewrite it, is most of what evaluating a strategy should mean.

→ See how shadow mode tests your rules for free · See what Evaluate keeps track of