What is mistake analysis, and why do people get it wrong?
Mistake analysis is a structured review of what you did, what you expected, what actually happened, and what you can change next time. It focuses on decisions and processes, not on trying to predict the future with the past. Common mistakes happen when the “analysis” part becomes vague, emotional, or outcome-only.
A frequent misunderstanding is to equate mistake analysis with “finding the reason it went wrong” without defining the decision and the measurement. Another is to analyze results as if they were guaranteed consequences of a single error. Markets, execution, fees, and changing conditions mean one outcome rarely has one cause.
Common mistakes and how they affect learning
1) Mixing the error with the outcome
A neutral mistake analysis separates the action (the decision or process) from the result (profit or loss, or another observed outcome). If you only label the outcome as “bad” and then retrofit a story, you may learn nothing actionable.
Neutral check: write the exact step you would change—then test whether that step was truly the causal factor, given your conditions and assumptions.
2) Skipping assumptions behind any example
When you show an example—such as “if spread were lower, the trade would have worked”—the calculation depends on assumptions (costs, timing, execution quality, and what “worked” means). Without stated assumptions, the example cannot be verified.
Neutral check: list every assumption explicitly, including what you are treating as fixed versus variable.
3) Treating stable mechanics as if they were stable outcomes
Some relationships are structural (for example, how costs reduce returns), but the future path of price is variable. Mistake analysis goes wrong when stable mechanics are used to argue that similar future events will produce similar outcomes.
Neutral check: separate “mechanics you can control” (process quality, documentation, consistency) from “conditions you cannot fully control” (market movement, execution details, external events).
4) Failure mode: confirmation bias during the review
A review can become a way to protect an earlier belief—only searching for explanations that support your thesis. This can hide other contributors like timing, sizing, or cost impact.
Red flags (rode vlaggen): the review only finds one explanation; it avoids disconfirming evidence; it never updates the model even when evidence contradicts it.
5) Using the review as a single-step “fix”
Mistake analysis often fails when it ends with one proposed change, without checking whether the change addresses the actual failure mode. If your process change is not connected to the defined error, you may repeat the same pattern under different conditions.
Klaarcriterium (ready criterion): you should be able to state what you will do differently next time and what evidence would show the change is working.
Evidence and a worked example of neutral checks (without claiming prediction)
Consider a review where a decision led to an unfavorable net result after costs. A common mistake is to conclude “the market moved against me” while ignoring whether the process allowed the cost impact to be too large for the expected edge.
A more neutral approach:
- Define the decision moment: what you based the decision on, and what you planned.
- Define the measurement: what you mean by “unfavorable” (net after costs, relative to a benchmark, etc.).
- List assumptions: cost estimates, timing assumptions, and what execution quality you are assuming.
- Check competing explanations: could the issue be timing, sizing, cost sensitivity, or an inconsistency with your own rules?
- Record one controllable change that targets the defined failure mode.
This does not claim future accuracy. It improves your ability to verify the reasoning you used.
Limitations and risks of mistake analysis
Mistake analysis is only as good as its inputs and its willingness to be specific. It can mislead when:
- you treat historical cause-and-effect as a guarantee for the future;
- you omit costs and execution assumptions from examples;
- you confuse emotional certainty with evidence;
- you fail to separate mechanics from variable conditions.
Outcomes also vary with market conditions, costs, execution, and jurisdiction, so even careful reviews cannot guarantee the next result will be better. The best you can do is create a review process that can be checked and improved.