How to use a poker solver without lying to yourself

1. Reconstruct the real decision point

Enter the board, pot before the action, remaining stacks, position, and legal action history. A wrong pot or actor changes every downstream value.

Use cumulative wager-to amounts consistently and distinguish a full raise from a short all-in.

2. State what is known and what is assumed

Your own cards and public board may be known. The opponent’s range usually is not. Label whether the range comes from a preflop model, observed actions, population evidence, or a scenario.

Do not interpret an automatic legal-combination baseline as a detected opponent range.

3. Bound the action tree deliberately

Every permitted bet size and re-raise depth changes the game. A browser model may need fewer sizes or representative ranges to remain inside a declared memory budget.

That abstraction is not a defect when it is visible, tested, and kept separate from exact equity. It becomes misleading only when hidden.

4. Check convergence and robustness

Exploitability describes deviation inside the solved finite game. It does not measure the distance to unrestricted no-limit Hold’em.

Compare nearby bet grids, range reductions, and iteration budgets. A recommendation that flips under small modeling changes is not robust enough to present confidently.

5. Review the result as conditional evidence

Ask what is exact, what is approximate, what data entered the model, and what would change the conclusion. Save the configuration and engine revision alongside any screenshot or article.