Curiosity, with the workings shown

Good question.
Let’s test it.

Everyone has a favourite piece of mahjong advice. We turn specific habits into testable comparisons, then show the result—including when the answer is “we don’t know yet”.

September 2026 · exploratory batch

No clear winner.
Still worth understanding.

Every reported 95% interval in this batch crosses zero. These runs have not established a reliable advantage for either policy. Explore the comparisons below to see what we tested and which questions remain open.

How to read a simulation result →
10
focused comparisons
4,864
simulated hands
Data + method
available with the reports

Choose a question

What happens if…?

Earlier research notes

Archived summaries use different experiment designs. They do not include the same uncertainty reporting as the September matched-deal batch. Treat their claims as exploratory, not established strategy.

Browse 9 earlier reports
At your next table

Practical strategy →

Work on shape, value and judgement for your ruleset.

Check the foundation

Sources & standards →

How learning, reference and experiments fit together.

A question worth asking

Suggest a study →

Have a question about a call, a discard or a scoring choice? Tell us what you’d like to explore.