Trust / conversion
How We Decide Something Is Actually a Poker Leak
A leak is declared only after evidence, opportunities, and confidence are on the table. Insufficient evidence is a valid result.
1. Parse the hand correctly
This step stays in the domain contract so a future analyzer can implement it without changing the site.
2. Normalize the decision context
This step stays in the domain contract so a future analyzer can implement it without changing the site.
3. Detect candidate patterns
This step stays in the domain contract so a future analyzer can implement it without changing the site.
4. Count relevant opportunities
This step stays in the domain contract so a future analyzer can implement it without changing the site.
5. Check sample reliability
This step stays in the domain contract so a future analyzer can implement it without changing the site.
6. Inspect decision composition
This step stays in the domain contract so a future analyzer can implement it without changing the site.
7. Compare with strategy evidence
This step stays in the domain contract so a future analyzer can implement it without changing the site.
8. Estimate potential impact
This step stays in the domain contract so a future analyzer can implement it without changing the site.
9. Assign confidence
This step stays in the domain contract so a future analyzer can implement it without changing the site.
10. Prioritize by value of fixing
This step stays in the domain contract so a future analyzer can implement it without changing the site.
11. Re-test on future hands
This step stays in the domain contract so a future analyzer can implement it without changing the site.
What AI does — and what it doesn't
AI explains evidence. It does not invent hand actions, stacks, samples, benchmarks, or solver outputs.
Questions
- Does the AI invent solver outputs?
- No. AI may explain evidence that already exists. It must not invent actions, stacks, samples, benchmarks, or solver numbers.
- What if the sample is too small?
- The review should say insufficient evidence and refuse a fake conclusion.
See the methodology applied to a sample.
Use the database you already built. Post-session analysis only.
See sample analysis