Here is a scene that happens at every American table, several times an afternoon.
A player picks a hand in the Charleston. They need four tiles. Someone across the table exposes a pung. Then a kong. The discards start coming faster. And the player keeps drawing, keeps discarding toward the hand they chose twenty minutes ago, and eventually throws the tile that wins it for somebody else.
They did not lose because they misread the card. They lost because they never reconsidered whether they were still in the hand.
Japanese mahjong has a word for the decision they skipped: 押し引き (oshi-hiki), push and pull. It is the single most discussed idea in the professional literature there, and it is almost entirely absent from how American players talk about their own game.
The question
At any point in a hand, there are only two honest options:
- Push — continue playing to win, accepting that you may deal into someone.
- Fold — stop trying to win and start throwing only tiles that cannot hurt you.
Most American players recognise these as things that happen to them. The improvement is to treat them as a choice you make, at a specific moment, on evidence.
The four inputs
You do not need to calculate anything. You need four reads, and you can hold all four in your head.
1. How far am I from ready? Count the tiles you still need. Not the tiles you would like — the tiles the card line actually requires.
2. Are those tiles still live? If you need a pair of red dragons and two are already on the table, you need the other two and there are two left in roughly forty unseen tiles. If you need three 5-dots and nobody has discarded one, you have four live tiles. This is the input beginners skip entirely, and it is the one that matters most.
3. Is anyone else close? Count exposed groups. A player with two exposed groups is close. A player with three is very close. A player who has exposed nothing is either concealed — dangerous and invisible — or nowhere.
4. What will it cost me if I am wrong? This is where the American game differs from riichi and where most players get the arithmetic backwards. In riichi, dealing into a big hand is catastrophic and folding is cheap. In the American game, being wrong costs you one hand value from each player — meaningful, but bounded.
The judgement
Put them together and the rule is simple:
Push if you need one or two tiles and they are plausibly live. You are genuinely in the hand. Take the risk.
Push if nobody at the table looks close. An empty table is an opportunity, even with a mediocre hand.
**Fold if you need three or more tiles and someone has exposed two or more groups.** You are not going to get there in time, and the tiles you need are exactly the tiles they are likely to want.
Fold if your hand is dead and you know it. This is the one that hurts. If the tile you needed was discarded three rounds ago and you did not call it, you are not playing any more. Stop pretending.
What folding actually looks like
Folding does not mean announcing anything. It does not mean throwing the hand in. It means one concrete change: you stop discarding toward your hand and start discarding by safety.
You keep your rack — you may get lucky, and you should still take a win if it falls on you — but every discard from that point is chosen because it cannot hurt you, not because it improves you.
The safest tiles, in order:
- Genbutsu (現物) — a tile the player you are worried about has discarded themselves. They cannot be waiting on it, because if they were, they would have won.
- A tile discarded twice by anyone. Two players have said they do not want it.
- A tile that cannot complete any group you can see. If every exposed group at the table is dots and dragons, a 3-bam is close to free.
- A suit nobody is collecting. If three players have been discarding cracks all afternoon, cracks are safe.
Note what is not on that list: the tile that is worst for your own hand. That is the discard most players make, and it is the one that ends hands.
The cultural caveat
One honest note, because this is a social game played in living rooms.
Fold too readily and the game gets worse for everyone. Three players throwing safe tiles at each other for twenty minutes is a tedious hand, and at most American tables it is also a breach of the unwritten contract — you are there to play, not to optimise.
The right balance: play your hand until the evidence says otherwise, then fold without announcing it. Folding is a private decision, not a statement. Nobody needs to know you have given up, and saying so out loud tends to make the table colder.
The one-sentence version
Most hands are not lost by the player who picks the wrong hand on the card. They are lost by the player who picked a reasonable hand, watched the table change, and never went back to look at it again.