When to Commit to a Big Hand — and When to Fold
Two rules: decide within the first three turns and never switch to a big hand after that; and abandon the moment a tile you still need is provably gone — all four copies visible across discards, melds and your own hand.
The commit thresholds
These are the actual entry conditions our AI engine uses. Each hand has two gates: a turn deadline for the decision, and a tile threshold your hand must already meet. Below the threshold it does not start — that discipline is itself the strategy.
Note how narrow the window is for Nine Gates and Thirteen Orphans — turn one or two. Those hands rebuild your entire holding, so a turn's delay is fatal. Small Dragons and Seven Pairs get three turns because they still allow mid-course correction.
The abandon signals
Committing is not the same as holding on. Any one of these signals means abandon immediately:
- A required tile is dead — all four copies of a tile you still need are visible across the discards, exposed melds and your own hand. It can never be drawn; the hand is over.
- You claimed the wrong meld — for Great Dragons, Great Winds and the dragon-suit hands, one off-target claimed meld kills the hand permanently, because that meld slot is locked.
- A concealed hand made any claim — Thirteen Orphans, Nine Gates and Seven Pairs are disqualified by a single chow, pung or open kong.
- Not enough pairs remain reachable — for Seven Pairs, if pairs already held plus singles that could still pair total fewer than seven, it is arithmetically finished.
- Nine Gates deficit exceeds supply — count the shortfall rank by rank against the 3-1-1-1-1-1-1-1-3 requirement; if any shortfall exceeds the copies still unseen, stop.
The underlying principle
Every abandon signal is the same idea: count what is visible rather than hoping. Each tile has four copies, and every copy in the discards, in any exposed meld, or in your own hand is accounted for. What remains is a definite number, not a feeling — and once it reaches zero, waiting costs you tempo and exposes you to dealing in. Strong play is not about chasing fewer big hands; it is about abandoning them faster.
What chasing actually costs
Chasing is never free: you trade away speed and defensive flexibility. Concealed hands make this starkest — everything must be self-drawn, which typically costs several extra turns to reach a ready hand, during which someone else may simply win. That is why the entry thresholds are strict: not because big hands aren't worth it, but because chasing below the threshold burns turns for nothing. For how many players actually complete each hand, see the special hand rarity data.
Frequently asked questions
When should you go for a big hand in mahjong?
Decide in the first one to three turns. Our engine uses hard thresholds: Seven Pairs needs four pairs already in hand before it commits; Great Dragons needs all three dragons present and seven or more dragon tiles. Below the threshold it plays a normal hand. Switching to a big hand mid-game is almost always too late.
When should you abandon a big hand?
The instant a required tile is dead. Count every copy visible across the discards, all exposed melds and your own hand — if all four copies of a tile you still need have appeared, the hand is unreachable and holding on only loses tempo. Claiming one wrong meld also kills most big hands permanently.
Big hand or fast win — which is better?
It depends on your starting tiles and the minimum faan. At a 3-faan minimum, All Chows plus self-draw and a concealed hand clears the bar without any big-hand risk. But when your deal already meets a commit threshold, chasing has much better expected value. The costly mistake is switching halfway — a hand caught between two plans wins neither.
Practise this
These rules are easy to read and hard to apply. In training mode the AI opponents run this same logic, so you can test when to commit and when to fold, with a post-game review of every decision.