Tile efficiency is the habit of choosing the discard that leaves the largest number of useful tiles still live in the wall. It is arithmetic rather than instinct: every hand has a countable distance from a win, and a countable number of tiles that shorten it. Three disciplines produce most of the gain — build exactly five blocks, prefer shapes that eight tiles can complete over shapes that only four can, and stop calling melds you do not need.
The two numbers that decide every discard
Shanten is how many tile exchanges separate your hand from ready. A hand waiting on its final tile sits at zero, a hand one exchange away is one-shanten, and a fresh deal usually starts at three or four.
The word is borrowed from Japanese riichi, where the ready hand is called tenpai; Wikipedia’s article on Japanese mahjong lays out that vocabulary if the terms are new to you. The counting itself works in any variant built from runs and triplets.
Acceptance is the count of tiles that would move you a step closer if you drew them. A discard that takes your acceptance from twenty-four tiles to twelve has halved your speed, whatever it felt like.
Because a standard set contains exactly four copies of every suit tile and every honour tile, both numbers have hard ceilings. One specific tile can never be worth more than four; a shape that two different tiles can complete is worth eight. Most of tile efficiency is refusing to settle for four when eight is available.
Build five blocks, not thirteen tiles
A winning hand is four sets plus a pair, five blocks, no more. The thirteen tiles in your hand should be five blocks under construction, and everything else is dead weight occupying a slot.
Newer players collect six or seven promising fragments, then spend the middle of the hand dismantling them, and every surplus block is a tile that could have been serving a block that mattered.
Take this thirteen-tile hand — 1-2-3 Bamboo, 5-6 Bamboo, 2-3 Circles, 8-9 Circles, 7-8 Characters, and a pair of 4 Characters. That is six blocks competing for five slots, so one of them has to die.
| Block | Shape | Completed by | Copies live |
|---|---|---|---|
| 1-2-3 Bamboo | Finished chow | Already done | — |
| 5-6 Bamboo | Two-sided | 4 or 7 Bamboo | 8 |
| 2-3 Circles | Two-sided | 1 or 4 Circles | 8 |
| 7-8 Characters | Two-sided | 6 or 9 Characters | 8 |
| 4-4 Characters | Pair | 4 Characters | 2 |
| 8-9 Circles | Edge | 7 Circles only | 4 |
The edge shape is the obvious casualty. Throw the 9 Circles now and the 8 Circles a turn later (or sit on the 8 for a round as a safe tile if the table is heating up) and the hand keeps three eight-tile shapes plus its pair, roughly two dozen tiles that genuinely advance it.
Holding the 8-9 Circles instead adds four tiles of acceptance and guarantees you will have to break something better within a few turns.
Waits are not created equal
The shape you finish on decides how often the win actually arrives. These are the five you will meet, counted with nothing yet visible on the table.
| Wait | Example | Wins on | Copies live | Verdict |
|---|---|---|---|---|
| Two-sided (ryanmen) | 4-5 Bamboo | 3 or 6 Bamboo | 8 | The default. Aim to finish here. |
| Closed wait (kanchan) | 4-6 Bamboo | 5 Bamboo | 4 | Fine when it glues two strong blocks together. |
| Edge wait (penchan) | 1-2 Bamboo | 3 Bamboo | 4 | Upgrade at the first opportunity. |
| Dual pair (shanpon) | 3-3 Bamboo plus West-West | 3 Bamboo or West | 4 | Worth it when the triplet adds real value. |
| Single tile (tanki) | A lone 5 Characters | 5 Characters | 3 | Late-hand tool; you hold one of the four yourself. |
Subtract everything you can already see
Those figures are ceilings, not facts. Every copy lying face up in a discard row or exposed in someone’s meld comes straight off the count.
A two-sided wait on 3 and 6 Bamboo begins as an eight-tile wait, but if three 3 Bamboo are already lying in the discards it is down to five live tiles, and once the fourth goes it is a four-tile wait wearing an eight-tile costume. Reading the pond before committing to a wait is the most valuable habit at intermediate level.
Upgrade weak shapes instead of freezing them
Weak blocks rarely improve themselves; the upgrade usually arrives somewhere else in the hand. If you are stuck with 1-2 Bamboo and then draw a 6 Circles next to your loose 7 Circles, the new two-sided shape is what finally lets the edge wait go.
This is why loose tiles in the middle of a suit are worth more than terminals: a floating 5 turns into a two-sided shape the moment a 4 or a 6 arrives, and into a closed wait with a 3 or a 7, so sixteen different tiles improve it. A floating 1 has no two-sided future at all: a 2 alongside it makes an edge wait and a 3 makes a closed wait, and both of those top out at four tiles.
A lone wind or dragon starts with three partners live, and every turn it sits unpaired the chance of it becoming a triplet falls. Keep it early for value or safety, then release it around the middle of the hand.
Calling is a trade, and most players overpay
Claiming a discard buys speed. It is paid for in three currencies, and beginners tend to spend all three without noticing.
- Information. An exposed chow tells three opponents which suit you are collecting and roughly where in it you are working.
- Flexibility. A concealed set can be broken up and repurposed later. An exposed one is welded in place for the rest of the hand.
- Value. Most scoring systems pay extra for concealed hands, and many Hong Kong tables agree a minimum score before a win may be declared at all. A hand of three called chows can be complete and still unpayable.
A workable filter: call when the tile takes you to ready or one exchange away, when it is the last live copy of something you need, or when your hand already clears the table minimum. Turn-three chows claimed to feel productive fail all three tests, and the mechanics of who may claim what sit in our walkthrough of how to start and play a mahjong game.
Efficiency has an off switch
Every count above optimises one thing: the chance your own hand finishes. None of it measures what happens when an opponent wins on a tile you handed them.
Once a player has two exposed melds and starts discarding honour tiles late, the efficient discard and the correct discard stop being the same tile. Our breakdown of when to play safe in mahjong covers the moment the count should be abandoned. The rough rule is to stay efficient while you are within two exchanges of ready and nobody has shown strength, and fold once you are three or more away while someone else is visibly close.
Five habits that quietly lower your win rate
- Carrying six blocks past the early turns. Decide which fragment dies while it is still cheap.
- Keeping a lone honour tile all hand. Its value decays every turn it fails to find a partner.
- Committing to a big hand on the first draw. Let the tiles you actually draw choose the target.
- Discarding the pair candidate. Every standard hand needs eyes, and pairs are harder to rebuild than runs.
- Never reading the pond. The discards tell you which of your waits are already dead.
How to train the counting habit
Say the numbers out loud for a few sessions. Before each discard, name your five blocks and the tiles that complete each one; it feels laborious for two games and turns automatic by about the fifth.
Volume is what fixes counting, and several apps surface efficiency statistics after a match — we compare them in our roundup of the best online mahjong platforms. Bear in mind that every shape above assumes a chow-based game; in the American card system the patterns are dictated by the card the National Mah Jongg League reissues every year, as explained in our guide to western mahjong rules.
Counting under mild social pressure is the version that transfers, and three people waiting on your discard supplies the pressure. Sydney players can hire the whole rig for a night (144 tiles, a table if the kitchen bench is too small, stools if the chairs are wrong) work through a dozen hands out loud, and hand it all back the next day.
Counting questions that come up mid-hand
What does shanten actually mean?
It is the number of tile exchanges between your current hand and a ready hand. Zero means you are waiting on your winning tile, one means a single useful draw away from waiting. Tracking it converts vague hand-feel into a number you can act on.
Is a two-sided wait always better than a closed wait?
No. A closed wait that completes a high-scoring hand beats a two-sided wait worth nothing at a table with a minimum score, and safety can override both. Where value and risk are equal, though, take the eight tiles over the four.
When is it right to keep an edge wait?
Keep it when it is your fifth block and nothing better exists, when the tile completing it is still fully live, or when it feeds a terminal-heavy pattern you are chasing for value. Otherwise it should be the first thing you drop when a stronger shape appears.
How much acceptance should a decent one-shanten hand have?
More than a dozen tiles is comfortable and anything in the twenties is strong, while single digits means you are one-shanten in name only. Rather than memorising thresholds, compare your candidate discards and take the larger number.
Does tile efficiency apply to American and western mahjong?
The counting logic transfers, the shapes do not. Because hands must match a pattern from the official card, efficiency there means committing early to the card lines your tiles support and dropping the ones the discards have already killed.
Is efficiency more important than reading opponents?
For your first hundred games, yes — most lost points come from hands that never finish rather than from clever opposition.
Learning on a proper table helps. Rent a complete mahjong set — tiles, table and racks, delivered ready to play across Sydney.
More in Beginner's Guides, Strategy, Tips & Tricks