Mahjong Discard Strategy: Choosing the Right Tile Every Turn

One Decision, Every Turn

Mahjong hands are won and lost at the discard. You make the decision every single turn, roughly eighteen times per hand, and each one either improves your position or quietly damages it. It is the most repeated decision in the game and therefore the one worth sharpening first. See also Mahjong Strategy: How to Actually Win More Hands.

The decision looks simple — give away one tile you do not need. The difficulty is that you usually have two or three candidates that all look unhelpful, and they are not equivalent. They differ in how much they could help you later, and in how dangerous they are to the other three players. See also Mahjong Scoring System Explained: Faan, Han and Points.

That is the shape of the problem. A discard is judged on two axes at once: its value to your own hand if kept, and its danger to you if thrown. Most of this guide is about evaluating those two axes separately before combining them.

Beginners tend to judge only the first axis, and often badly. They throw the tile that looks least useful right now, without asking what it could become in three turns or what throwing it tells the table. Both omissions are correctable, and fixing them changes results faster than almost any other adjustment.

There is also a tempo element that makes the discard decision unusual. Unlike most choices in a game, you cannot defer it. You must throw something every turn, and the option of waiting for more information does not exist. That means your discard routine has to work even when you are uncertain, which is why the later sections of this guide spend so much time on defaults and tiebreakers rather than on ideal cases.

Comparing Candidate Discards

When you have two or three candidates, the comparison is about what each tile can still become. A tile that can join several different blocks is worth more than one that can join a single specific block, even if both look equally lonely in your rack.

Take a concrete case. You hold 1 and 5 in a suit, and you must discard one. The 1 can only be completed toward 1-2-3, so it needs the 2 and 3 to arrive. The 5 sits in the middle of the suit and can form a sequence with 3-4, 4-6, or 6-7. The 5 is worth substantially more, and if you must choose, the 1 goes first.

The same logic applies to blocks rather than single tiles. Given two partial blocks and the need to cut one, the block that accepts more completing tiles survives. A two-sided block such as 2-3 accepts tiles from both ends. A closed block such as 1-2 accepts from one end only. Where the rest of the hand is equal, keep the two-sided shape.

A worked comparison shows how much this matters. Two hands hold the same nine completed tiles. One has a 2-3 block, the other has a 1-3 gap block. Both look like a partial block of the same size. But the 2-3 accepts two tile types from two directions, while the 1-3 accepts one tile type from the middle. Given a choice, the 2-3 survives every time, and the difference is not subtle — it is roughly double the ways to complete.

Compound shapes deserve special care because they are easy to break by accident. Tiles like 2-2-3-3 look redundant, and a beginner discarding one of them is destroying a shape worth roughly double the equivalent separate blocks. Before discarding from an overlapping run, check whether the overlap is what gives the shape its flexibility.

The exception to all of this is a tile whose completing tiles are visibly gone. Keeping a shape that needs tiles already in the discard pile is not flexibility, it is a dead branch. Count visible copies before you commit to keeping a block.

A second exception is a tile that is far from the rest of your hand in suit. A beautiful two-sided block in a suit where you hold nothing else is worth less than a mediocre block that connects to tiles you already own, because blocks only become sets if they can be joined by the sequences around them. Flexibility measured in isolation is not the same as flexibility in your actual hand.

Counting Live Copies

Four copies of every tile exist, and each one you can see reduces what remains. This turns discard choice from an abstract judgement into a countable one.

Suppose your hand’s remaining path runs through tile X, and you can see three of the four copies of X on the table. One copy remains in play, possibly in the wall and possibly in an opponent’s hand. Your path is nearly dead. That is a strong argument to take a different route, even if the route currently looks worse.

The reverse also holds and is more motivating. If none of the copies of the tiles your block needs are visible, that block is at full strength and deserves protection. Holding it is not optimism; it is bookkeeping.

Where this becomes decisive is when comparing two shapes that look equivalent. Both might accept four tile types, but if one set of four is heavily visible and the other is not, the effective value differs sharply. The visible tiles are not just gone for you; their absence tells you the shape is less likely to complete.

The habit to build is narrow. Do not count all thirty-four tile types. Count the specific tiles your one or two live blocks depend on. That is a small enough set to track while the hand develops, and it catches most of the cases where a shape is secretly dead.

There is a useful way to convert this into a decision. When you are choosing between keeping block A and block B, count the visible copies of every tile that would complete each. Whichever block has more live outs is the one to keep, even if it looked weaker before you counted. This is the single most reliable tiebreaker available, and it takes a few seconds.

一筒 · 1 Dot四筒 · 4 Dot七筒 · 7 Dot中 · Red Dragon中 · Red Dragon

The Discard Pile as a Threat Map

Your discards are a decision. Everyone else’s discards are data. The pile in front of each player is the clearest public information in the game, and it should be running as a background process on every turn.

Three readings matter most. First, concentration: a player discarding one suit while leaving another untouched is building in the untouched suit. Second, timing: an early discard is nearly meaningless, while the same tile thrown after ten turns says the player built around it and did not want it. Third, commitment: a player who has called several tiles has a constrained hand, and their later discards are unusually informative.

The direct use of all this is safety. A tile that appears in the discard pile of the player you are worried about is the safest tile you can throw, because they cannot win on their own discard. That check belongs in your discard routine, before any aesthetic judgement about which tile you would rather lose.

The indirect use is predicting. If a player has discarded 4 and 6 in a suit, the 5 is unlikely to be in their hand, which means your own 5 is safer than its central position suggests. This is the gap principle, and applying it is how a player moves from avoiding obvious mistakes to actively steering.

A practical simplification: track one opponent properly rather than three badly. The one who has called tiles and is discarding narrowly is most likely to win, so their discard pile is the one worth memorising.

The reason the discard pile beats every other information source is that it is authoritative. Body language, timing and hesitation are suggestive but can be faked or misread. A tile on the table is a fact: that player chose not to keep it, and in most rulesets cannot win on it. Building your defensive picture on facts rather than inferences is what makes defence reliable rather than speculative.

When Two Discards Are Equal

Often the comparison produces a tie. Two tiles help your hand equally, or the difference is too small to resolve. In that case danger decides, and danger is easier to read than value.

A tile already in the threatening player’s discard pile wins any tie. If neither is there, prefer the tile with more copies visible on the table. If it is still tied, prefer the tile that is further from the middle of its suit — terminals and honours are generally safer than central tiles because fewer waiting shapes need them.

There is a fourth tiebreaker that beginners never consider: the information you leak. Discarding from a suit you have been avoiding tells the table something about your hand. Discarding a tile that keeps your options ambiguous leaks less. When everything else is equal, prefer the discard that reveals least.

This matters more late in the hand. Early on, no opponent can use your discard pattern meaningfully, so the information cost is near zero. In the final few turns, a single revealing discard can tell a strong opponent exactly which tiles you need, and they will hold them.

The order of tiebreakers is worth memorising because it removes hesitation under time pressure: safety against the tracked opponent, then visible copies, then distance from the middle, then information leaked. Applying them in that order costs a second and prevents most rushed errors.

One refinement that matters in the last few turns: when the game is nearly over, information leaked becomes much more valuable to your opponents than it was early on. A single discard that reveals which suit you are waiting on can let a strong opponent hold the exact tile you need and win the race. In the final turns, prefer the discard that says least, even if it is marginally safer to throw a different tile.

A Discard Routine

The mechanics above only help if they run every turn without requiring deliberation. So the deliverable of this guide is a routine, short enough to execute reliably at speed.

List your candidates. Usually two or three tiles in your hand are genuinely discardable this turn. Identifying them explicitly prevents the common failure of discarding the first tile that looks wrong.

Compare them on what they could become. A tile that can join more blocks survives. A compound shape is protected. A shape whose completing tiles are mostly visible loses its claim.

Check the tracked opponent’s discard pile. Any candidate appearing there is immediately the best choice, because safety against a committed player outweighs small differences in value.

If the candidates are still tied, apply the tiebreaker order — visible copies, distance from the middle, information leaked.

Then decide, and move on. The point of a routine is speed; a hand played well is not a hand played slowly. The comparison should take a few seconds once the habit forms, and the improvement comes from making the right choice consistently rather than from agonising over individual turns.

The reason this pays off across a session rather than only in dramatic hands is that discards compound. Each correct discard preserves one more option in a hand that needs many. Eighteen small correct decisions per hand do not produce a highlight, but they are how hands that looked unremarkable keep reaching tenpai.