Two entirely different games answer to the name mahjong, and advice written for one is close to useless for the other. The first is a solitaire puzzle: a stack of 144 tiles arranged in some shape, stripped away in matched pairs until the table is bare or nothing legal remains. The second is the four-player draw-and-discard game the tiles were cut for, where you assemble four sets and a pair while three opponents feed you as little as they can. Geometry and removal order decide the first. Information and price decide the second. This article keeps them in separate halves rather than blending them into one list of platitudes.
Part one: how a solitaire layout actually works
A layout is a dependency graph pretending to be a decoration. One rule generates all of it: a tile can be lifted when nothing rests on any part of it and at least one of its long sides, left or right, has no neighbour. The turtle, the spider, the pyramid, the fortress and the cathedral are the same puzzle wearing different masks. In each of them a small number of tiles are load-bearing, most are not, and the load-bearing ones decide whether you finish.
Which means the first useful thing to do on a fresh board is not to look for a match. It is to find the worst-placed tiles: the ones under two or three layers with neighbours pressing on both sides. Those tiles are on a deadline, because the only thing that frees them is the removal of what sits above, and a layout never gains extra pairs once it has been dealt. From that moment every match should either shorten the queue above one of them or be a pair you can genuinely afford to lose.
The arithmetic of four
Nearly every design in the set exists in four copies. The three suits run one to nine with four of each, and there are four of every wind and every dragon. The eight flowers and seasons are the exception: most solitaire versions treat them as two loose groups of four, so any flower answers any other flower. You are therefore never looking at a tile and its match. You are looking at one of four, and choosing which of the remaining three to spend it on.
That choice is where boards are won and lost. If all four copies of the six of bamboo are liftable, clearing any two of them costs nothing: the other two remain a legal pair wherever they sit. If two are liftable and the other two are stacked one directly on top of the other, taking the easy pair kills that design outright. The covered copy cannot be raised while its twin sits on it, the twin has nothing left to match, and both go on pinning whatever lies beneath them for the rest of the game.
The habit worth drilling: when three or four copies of a design are in play, work out the pairing that clears all four rather than the pairing nearest the cursor. The way boards run out of moves is almost always a variation on that single error.
Depth first, rim last
Beginners clear the outside, because the outside is where the free tiles are. It feels like progress and it is the most reliable way to lose. Rim tiles were never the constraint: they had open edges when the board was dealt and they will still have open edges ten minutes later. The tiles under the middle of the stack are the ones under pressure, and the supply of pairs available to dismantle the stack above them is finite. 10 Mahjong is one of the layouts kept on this site.
The practical form of the same idea: given two legal matches, take the one higher up the stack over the one further out on the rim, and take the one that uncovers something over the one that uncovers nothing. The full step-by-step version of that order is set out in the removal order that finishes boards.
Free edges are the currency
It helps to treat the board as holding a stock of legal moves that every removal either spends or mints. Lifting a tile from the end of a row mints one, because its neighbour inherits the open edge. Lifting a tile off the top of a stack mints whatever the tile below can use: two new options sometimes, none at all if that tile is walled in on both sides. Lifting the single tile that joins two clumps of the layout mints nothing and quietly narrows the board.
Spare tiles belong to the same budget. A liftable tile that covers nothing and blocks nothing is not clutter, it is a partner held in reserve for something buried. Spend the reserves early on tidy little matches and the endgame arrives with plenty of free tiles and no two of them alike.
Choosing between two legal matches
When several matches are available at once, this order settles most of them:
- Uncover before unblock — a match that takes a tile off the top of a stack beats one that merely shortens a row.
- Deep before shallow — the higher the layer, the more weight it is carrying.
- All four before any two — a pairing that clears every copy of a design costs you no future options at all.
- Busy before empty — work the crowded half while the quiet half still holds spare partners.
- Never split a working pair — two copies that are the only keys to a locked third are not a match, they are tools.
What a shuffle really costs
A shuffle does not rebuild the layout. It redeals the faces into the same slots, so every blocked position stays exactly as blocked as it was and only the labels move. A shuffled board can therefore come back with no legal move at all, and often does when the geometry rather than the deal was the problem. In scored versions you pay twice over: most dock points, some take seconds off the clock, and nearly all reset whatever streak bonus was running.
There is a third cost that no scoreboard prints. Five minutes of play leaves you knowing roughly where the fourth copy of everything is, and a shuffle erases that in one click. Take one when the board is genuinely dead, not when you dislike the two moves you can see.
Hints and timers are not free either
A hint shows a legal match. It says nothing about whether that match is a good one, and on a board where three pairs are available it will happily point at the one that strands a design for the rest of the game. Its honest use is diagnostic: when the hint finds nothing, the board is dead and you have the answer without staring at it for another minute. Treat the button as a stall detector rather than as a source of moves.
Timed versions rearrange the priorities above rather than replacing them. Under a clock the cost of thinking is real, so the preference for deep tiles survives but the endgame arithmetic does not, since there is no time to solve twelve remaining tiles in your head while seconds drain away. The workable compromise is to think hard through the first third of the board, where the decisions are worth the most, and play by pattern after that. Untimed boards invert it: the opening is cheap, so put the attention where the stack is thickest and let the rim look after itself. A layout of that kind in the catalogue here is Gorillas Tiles Of The Unexpected.
Shapes change the tactics
The freeing rule is constant; the shape is not. Tall boards turn on the difference between the crown and the slab it stands on, which is why tall layouts need their own order of removal and why the tower layouts feel harsher than a turtle of identical tile count. Flat connect grids replace stacking with pathfinding and reward work along the border, covered in the connect-grid tactics next to the connect boards themselves. Rotating blocks hide a third of the tiles behind the camera, a different problem again: see rotating-cube tactics and the Dimensions layouts. For plain stacked shapes to practise the order on, the classic solitaire layouts are the place to start. The site’s own example of it is Mahjong 3D.
Part two: strategy at the four-player table
Now the other game. In the Chinese and Japanese families you hold thirteen tiles, draw a fourteenth on your turn and discard one, and you win by holding four sets plus a pair. A set is three of a kind, four of a kind, or a run of three in one suit. Winds and dragons cannot form runs, only triplets, and that single restriction shapes most of the decisions in the opening.
Pick the hand your tiles already point at
The usual beginner error is choosing a hand for its value and then trying to collect it. Do the reverse. Sort the thirteen tiles, count how many are already part of a pair, a triplet or a partial run, and work out how many draws separate you from a complete hand. That number, not the payout, is what to optimise in the first six turns. A hand three tiles from completion is worth more than a spectacular hand seven tiles away, because the wall runs out before the seventh draw arrives often enough to matter.
Two shapes to notice while sorting. A run missing its middle tile is weaker than it looks, since exactly one design completes it, while a run missing an end can be filled from either side. And pairs of terminals and honour tiles are cheap to hold in the first few turns and very expensive to draw in the last few.
Flowers and seasons, in the rule sets that use them, are set aside as they arrive and replaced from the wall, so they cost nothing to keep and never belong to the count of thirteen. Beginners regularly try to hold them as though they were suit tiles and play a turn short as a result.
The discards are the only public information
Everything you know about three hidden hands comes from what has been thrown away and how quickly. A player who has not discarded a single bamboo tile in ten turns is collecting bamboo. A player who throws a tile you were waiting on and then throws its neighbour has probably abandoned that suit. And once all four copies of a design are on the table or claimed, any wait on that design is dead, so count the copies before committing to a pair.
The same information runs the other way. Claiming a discard to complete a set is never free: it exposes the set to everyone and announces what you are building. Claim when it saves real turns, not to look busy.
Knowing when to take the cheap hand
A hand you never finish scores nothing, and that is the whole case for the cheap hand. When the wall is running down, or when the discards show that somebody else is a tile away, the expensive hand you have been nursing has stopped being an investment and become a liability. Finish something. A small completed hand also denies the round to whoever was about to collect a large one, which is worth considerably more than its face value.
Counting the wall turns that judgement into arithmetic. Work out roughly how many draws each player has left before the tiles run out, then compare it with how many tiles separate you from a complete hand. Two draws remaining and three tiles needed is not a hand, it is a hope, and the play is to switch to discarding safely and take the draw.
The defensive version is the same reasoning with the sign reversed. If you are four tiles away and the table has told you a neighbour is one away, stop feeding it: throw tiles that have already appeared in the discards and let the round end small.
What actually transfers between the two games
Two habits carry over, and only two. The first is counting copies, since there are four of everything and knowing how many have gone tells you what is still possible, whether the tile in question is buried under a stack or sitting in somebody else’s hand. The second is keeping your options for late rather than early: leaving spare partners on a solitaire board and holding a two-sided wait in a hand are the same instinct expressed in different games. Everything beyond those two belongs to the specific shape in front of you, which is why the tactical pieces in this section are split by board type rather than dressed up as universal rules. Mahjong Tower puts that layout on a playable board.
