What the Teachers Actually Disagree About
Five strategy questions with no settled answer, and five that are settled. Knowing which is which is worth more than any single answer.
On this page
The rules of this game have an authority: there is a rulebook, and where sources differ you can go and check. Strategy has nothing of the kind. What circulates as "consensus" is the overlap between roughly eight teachers, and the overlap is smaller than it looks.
So here is the map. Not our opinions — who holds which position, and where the ground is actually firm.
What is settled
Five claims survive independent checking across several teachers. These you can act on:
- Keep pairs. A joker cannot fill a pair, so pairs are the scarce resource.
- Do not pass a hand away. Watch what your three tiles add up to for whoever receives them.
- Exposing too early is the number one beginner error.
- Danger is measured by what is visible. Count copies on the table and in the discards.
- From an opponent's third exposure onward, stop feeding them.
What is not
Why the disagreement is the useful part
Two reasons, and neither is a dodge.
The first is practical. If you know a question is contested, you stop looking for the answer and start looking at your own situation — which is where the answer actually lives. "How many hands should I chase" has no general answer, but "can I hold three patterns in my head while watching three opponents" has one, and it is about you.
The second is defensive. A source that presents any of the five contested questions as settled has not read the others. That tells you something about how much weight to put on the rest of what it says — including its rules content, where being wrong is not a matter of taste.
There is a sixth contested tactic we have left out. It appears only in sources that get verifiable rules wrong elsewhere, and it contradicts every teacher above who has a track record worth anything.
A map of disagreement is not an obligation to repeat everything anybody has said. Where the only people advancing a position are people we can already show to be unreliable, that is not a disagreement between teachers — it is one side being wrong, and passing it on would launder it.
Three numbers that come from exactly one place each
These circulate as though they were established. Each appears in a single source, so use them as somebody's rule of thumb rather than a threshold:
- Do not expose before you have seven tiles toward a hand.
- The endgame begins around 38 tiles left in the wall.
- Stop passing when you are within four tiles of mahjong.
And one that expires every April
"Which tiles are precious this year" is a real question with a real answer, and the answer changes when the card does. 2019 favoured 5s; 2022 favoured 2s. Any claim of that shape without a year attached to it is either already wrong or about to be.
Serious strategy discussion happens in private groups no crawler reaches. What is mapped above is what is publicly checkable, and it is a fraction of what exists. More searching does not fix that — a table does.
Where the ground is firm
The rules, checked across eight sources — including the handful the sources get wrong.
Read the rules guidesWhere this comes from
Every position above is attributed to the teacher who holds it. The five settled claims survived checking across multiple independent teachers. Nothing here is our own strategy opinion, and none of these people is an authority in the way the rulebook is for rules.
The rules the settled claims rest on — that a joker may never sit in a pair, and that four copies of every tile exist — are confirmed by the American Mah-Jongg Association's rulebook and Tom Sloper's rules FAQ.