Anomaly probability testing
“He feels off” is not evidence — but it can become evidence. This lesson shows how to turn a long-run win rate into a p-value with a binomial test: not to decide “he is cheating”, but to answer “under a fair game, how likely is this data?”
🔒 Members only · Course
You are reading the free preview. Become a member to unlock all 7 lessons on this page, plus all 8 stages, the training tools and review templates. Payment is handled by our support team, with a 30-day no-questions refund.
Full system ¥299 — all 8 stages + 20 courses + training tools and review templates
Turn “something feels off” into a computable p-value with binomial and chi-squared tests: a win-rate significance checker, a dice fairness test, and a required-sample-size calculator — all computed locally.
What you will learn on this page
- 1. Why probability, not intuition
- 2. Win-rate checker
- 3. How to record so it matters
- 4. Dice distribution test (chi-squared)
Lesson 1 below is free to read · Lesson 2 onward is member content
All calculations run locally in your browser; no data is uploaded.
1. Why probability, not intuition
Suppose someone played 100 games and won 40. Intuition says “40% is high, something must be wrong”. But intuition makes two kinds of mistake:
- Underestimating luck — winning 35% over 100 games can happen purely by chance;
- Overestimating luck — winning 30% over 1000 games should not be explained by luck anymore.
The binomial test’s job is to turn “how much luck can explain” into a number, so the judgement no longer depends on feeling. The question it answers is: if the game is fair, how likely is it that this player’s win rate reaches exactly 40%?
🔒 End of the preview — the full text is for members
The rest of this lesson is sent only to accounts with access; it is not in the page source either.
This page includes
One payment, lifetime access; monthly new content is not charged again. Pair it with the free trainer on the homepage to practice together.
After this you can
- 1. Why probability, not intuition
- 2. Win-rate checker
- 3. How to record so it matters
- 4. Dice distribution test (chi-squared)
Lesson-by-lesson breakdown · 7 lessons on this page
Each lesson is one independently executable step: the principle, a hand example, and a checklist you can follow. Non-members can read Lesson 1; the rest unlock with membership.
1. Why probability, not intuition
Already readable2. Win-rate checker
3. How to record so it matters
4. Dice distribution test (chi-squared)
5. How many games to detect an anomalous win rate
6. Using statistics in the right place
7. Try it once
Sources
Rules follow the World Mahjong (MCR) competitive rules and mainstream Japanese riichi rules; methods and examples are compiled by this site. For terms and fan values see the glossary, yaku illustrated; differences in local variants (Sichuan, Cantonese, Changsha, etc.) follow the local table’s actual rules.