Start with the exact question
“Should I call?” is too broad for one number. A study might compare a player who calls eagerly with one who uses a particular decision policy. The opponents, rules and rest of the behaviour matter as much as the headline.
Read the tested change first. If the experiment never examined your situation, its average cannot answer that situation by itself.
Find the measure, then the uncertainty
Win rate and net points answer different questions. A player can win more hands while losing more points. Results in different scoring units cannot be ranked against each other just because their numbers look similar.
An interval crossing zero means the comparison has not established a direction at that level of uncertainty. It is not evidence that two approaches are identical. The September notes expose the sample sizes and paired outcomes so you can see that distinction.
Separate an experiment from a rule
Our Lab uses software players. Its outputs describe those policies and the model they played in; they do not amend an official rule or certify the best move against people. More simulated hands cannot repair a missing rule in the model.
Use research to sharpen a question and test a habit. Use the rulebook to settle legality. The guides and Lab reports help you explore both.
Saved in this browser. Completion is yours to mark when you feel ready.
Choose your next question →
Visit the learning library or practise at the table.
Sources & scope
An original Hé Fēn explanation with our own teaching examples. Reference links checked 13 September 2026. We are not affiliated with these organisations.
- September 2026 matched-deal comparisons Hé Fēn Lab. Original exploratory simulation data, experiment definitions and model limitations; not a human-player study.