Definition: what “mistake analysis” means
Mistake analysis is a structured process for reviewing a past decision (for example, a trading decision, model decision, or risk decision) to determine what specifically went wrong and why. It treats the outcome as evidence, not proof, and focuses on controllable inputs such as the quality of your assumptions, the consistency of your rules, and how execution and costs affected the realized result.
A “worked example” means the process is shown step by step with explicit numbers (or a fully specified scenario) so you can reproduce the reasoning. In practice, it usually includes:
- List the decision and its stated goal
- Record inputs and assumptions you used at the time
- Identify the mismatch between what you expected and what happened
- Attribute the mismatch to likely causes you can justify
- Quantify the effect of each cause as far as the available data allows
How it works: mechanism, inputs, and calculation structure
To keep mistake analysis verifiable, use a calculation skeleton with clearly labeled parts:
- Expected result (based on your own assumptions): what you thought would happen.
- Realized result (based on what actually occurred): the value you can compute from your records.
- Costs and frictions: spreads, commissions, funding, slippage, or other charges—only if you explicitly assume them.
- Attribution: split the difference between expected and realized into components you can support.
A key distinction is between:
- Stable mechanics: the way you compute profit/loss, apply a rule, or compute a difference.
- Variable conditions: market movement, liquidity changes, execution quality, or other factors outside your control.
A worked example should state assumptions for every numeric step. If an input is unknown, you either omit it or you assign a stated assumption range and label it clearly.
Worked numerical example (with explicit assumptions)
Scenario
Assume you planned a single long position with a fixed size and held it for the same time window. You can compute everything below from your own assumed inputs.
Given assumptions (recorded at the time)
- Position size: 10,000 units
- Entry price (when you decided): 1.1000
- Exit price (what actually happened): 1.1050
- Contract value conversion for this example: treat a 1 price-point move as producing 10,000 quote-currency units of value change (so we can compute profit directly from price difference).
- Estimated total costs (including spread/commission/slippage): 0.0002 in price terms.
Step 1: Compute expected move
Assume your expectation at entry was that the price would rise to 1.1040.
- Expected price change: 1.1040 − 1.1000 = 0.0040
Step 2: Compute expected profit before costs
Using the conversion assumption:
- Expected profit (before costs) = 0.0040 × 10,000 = 40.00
Step 3: Compute realized profit before costs
- Realized price change: 1.1050 − 1.1000 = 0.0050
- Realized profit (before costs) = 0.0050 × 10,000 = 50.00
Step 4: Apply assumed costs
Costs in this example are modeled as a price drag of 0.0002.
- Cost amount = 0.0002 × 10,000 = 2.00
So:
- Expected net profit = 40.00 − 2.00 = 38.00
- Realized net profit = 50.00 − 2.00 = 48.00
Step 5: Identify the mistake as an explanation of difference
The difference between realized net and expected net is:
- Mistake analysis target: 48.00 − 38.00 = 10.00
Now attribute likely causes, using only what your records support. Here is one transparent attribution that you can verify:
- Expectation error (primary): Your assumption about the exit level was too low (you expected 1.1040 but got 1.1050).
- Costs modeling error (secondary, only if true): You assumed total costs equal to 0.0002 in price terms.
Notice what this does and does not claim. It does not claim that the “mistake” was guaranteed to be the only cause. It simply shows a reproducible mapping from your assumptions to your result.
Material limitation / failure mode
A common failure mode is using the outcome to explain the past too confidently. Even with perfect arithmetic, you may mis-attribute the cause because you lacked information at the time (for example, the signals, news, or order-book conditions you did not observe).