Post-Tournament Audit: Process vs. Outcome
A tournament result answers what happened. It does not, by itself, answer whether the original decision used the information available at the time well. A useful audit preserves the pre-event thesis, price, and assumptions before looking for lessons in the final score.
This worksheet works for a bracket choice, a futures position, or an OwnTheLines simulation selection. It does not require a proprietary model or a large historical dataset. Its purpose is to make the next decision more consistent, not to turn one win or loss into proof of skill.
Preserve the Original Decision Record
Complete this section before reviewing the outcome:
| Field | What to record |
|---|---|
| Event and market | The exact game, side, total, future, or bracket choice |
| Decision time | Timestamp and information cutoff |
| Entry price | Point value and attached odds, or pool scoring cost |
| Estimated probability | Your estimate and how it was produced |
| Thesis | Two or three falsifiable reasons for the choice |
| Known uncertainty | Injuries, role changes, small samples, or missing data |
| Planned size | The pre-committed simulation unit and rule used |
If any field is reconstructed from memory after the result, label it as reconstructed. Otherwise hindsight can quietly rewrite what you believed.
Separate Information Available Then From Information Learned Later
Create two columns. The first contains only facts available when the decision was made. The second contains later information: a late injury, a rotation change, unexpected foul trouble, or a tactical adjustment revealed during the game.
Do not penalize the original process for unknowable events. Do flag information that was public and relevant but omitted. This distinction prevents both outcome bias—calling every loss a bad decision—and self-protection—calling every loss bad luck.
Classify the Error
- Data error: a recorded input was wrong or stale.
- Model or rule error: the calculation or decision rule did not represent the intended thesis.
- Interpretation error: the facts were correct but the conclusion did not follow.
- Execution error: the selected price, timing, or simulation size broke the stated plan.
- No identified process error: the result varied from a reasonable pre-event estimate.
Compare the Estimate With the Price
Expected value is only as reliable as the probability estimate supplied to it. For a transparent hypothetical example, suppose a simulation decision risks 1 unit to return 1.10 units of profit, with an estimated 55% win probability:
Illustrative EV = (0.55 × 1.10) − (0.45 × 1.00) = +0.155 units
The arithmetic is exact given those inputs; the 55% estimate is not. The audit should therefore ask how that estimate was formed, what range of plausible estimates would change the decision, and whether the same method was applied consistently.
Never mark a process “good” solely because the calculated EV was positive. An unsupported probability creates a precise-looking but unsupported result.
Define One Future Process Change
Finish the audit with one action that can be checked next time:
- Keep the original thesis unchanged in the record.
- Identify the most consequential error classification.
- Write one narrow rule change, such as requiring two independent injury confirmations or recording point value and price together.
- State when the change will be evaluated.
- Do not change multiple model weights because of one tournament.
An actionable conclusion is specific: “Record the source timestamp for every availability assumption.” “Be more careful” is not.
Audit Questions
Use Statistical Variance to interpret noisy results and Confirmation Bias to improve the pre-decision record.
Q: Should a losing selection always trigger a change?
A: No. Change the process when the review identifies an error or when a sufficiently large, relevant sample challenges an assumption—not merely because one result lost.
Q: What if I did not record an estimated probability?
A: Audit the evidence and rule you actually used. The first improvement may be to record a probability range before the next decision rather than inventing one afterward.
Q: Can I use this without a model?
A: Yes. A written thesis, information cutoff, price, uncertainty list, and pre-committed simulation size are enough to make the review useful.
Q: How often should I audit?
A: Use a short review after a defined group of decisions and a deeper review after a tournament or season. Choose the interval in advance so a dramatic result does not determine when you inspect the process.