Mahjong Defense: How to Stop Dealing Into Winning Hands

Not Dealing In Is a Skill

Most mahjong guides teach you to win. Almost none teach you to lose less, even though the second skill decides more sessions than the first. A player who wins one hand and deals into three has a worse night than a player who wins nothing and pays nothing. See also Mahjong Strategy: How to Actually Win More Hands.

The asymmetry that makes defence worth learning is in the scoring. When you win, you collect from one or sometimes all players, and the amount depends on your hand. When you deal in, you pay a large share — often the full value of a good hand that someone else built patiently. One deal-in can erase two wins. See also Common Mahjong Mistakes: 10 Errors That Cost You Hands.

So defence is not the absence of aggression. It is a separate decision that runs alongside your attacking plan, and it occasionally overrides it. This guide treats it that way: what makes a tile dangerous, how to choose a safe substitute, and when abandoning your hand is the correct play rather than a surrender.

The reason beginners struggle with defence is that it feels like giving up. Building a hand is visibly productive, and discarding safe tiles feels like treading water. But the goal was never to build a hand. It was to finish the session ahead, and sometimes those two goals point in different directions.

There is a measurable way to see the value, which makes the discipline easier to accept. Track a session and count what you paid out versus what you collected. Most beginners find that deal-in payments exceed their winnings even in sessions where they felt they played well. That single comparison is usually enough to convert a reluctant defender into a committed one.

Reading Danger From Discards

Every discard on the table is a clue about what its owner does and does not need, and the clues compound into a usable picture of how dangerous each tile is.

Start with suit concentration. A player who discards one suit repeatedly, especially into the late middle game, is not collecting that suit. If their discards are mostly one suit while they never touch another, they are building in the untouched one, and every tile you hold in that suit is a candidate danger.

Next, read call patterns. A player who has called two or three tiles has committed their hand publicly. Their discards from that point are constrained — they are throwing what their finished shape does not need. That constraint is information. A hand still fully concealed tells you much less.

Watch for the tiles a player will not throw. If someone has been discarding freely all game and suddenly starts throwing only tiles already visible on the table, they are playing defence themselves, which tells you their hand is either unfinished or abandoned and they are no longer a threat.

The strongest single signal is a late, deliberate discard. Early discards are noise — a player simply had an unwanted tile. A tile thrown after ten or more turns, when the player has had many chances to use it, says they built around it and did not want it. That says something real about their waiting tiles.

Reading all three opponents at once is beyond most players, and it is not necessary. Identify the one who has committed and is closest to completion. Defend against them specifically and treat the others as background threats.

One further read is worth adding because it is easy and accurate: watch how a player handles their discards physically or in the interface. A player who discards instantly every turn is playing by reflex and their discards carry less information than they appear to. A player who pauses before discarding, especially before throwing a tile that looks harmless, is choosing deliberately — which usually means they are weighing safety, and that they have something to protect.

Suji and the Gap Principle

The most useful defensive heuristic in mahjong comes from a structural fact: no two-sided wait can include a tile that sits between two tiles its owner has already discarded.

Here is the reasoning. A two-sided wait needs a block like 2-3, which accepts 1 and 4. For that player to be waiting on that block, they must hold 2 and 3. If they have already discarded the 4, they cannot be holding the 2-3 block, because the 4 would have completed a sequence they had. So a tile discarded by a player makes the tiles that would pair with it in a sequence — the ones two steps away — safer against that player.

This is what players call suji, and the pattern is worth seeing concretely. If a player has discarded 4, then 1 and 7 are safer against them than they look. If they have discarded 1, then 4 is safer. The tiles two steps away from a visible discard lose most of their danger from two-sided waits.

Two limits keep this honest. First, suji only protects against two-sided waits. A player waiting on a closed gap shape or a single tile is not bound by the pattern, and those waits ignore it entirely. Second, respecting suji too rigidly makes you predictable — a strong opponent notices which tiles you treat as safe and works backward to your hand.

Use suji as one input among several. It is a well-founded reason to consider a tile safer. It is not proof, and it never overrides a stronger signal such as a player calling a tile they already discarded.

A useful way to hold suji in your head without memorising tables: for any tile you hold, ask which partial block an opponent would need in order to be waiting on it. If the tiles that would have completed that block are already in their discard pile, the block is impossible and the wait is unlikely. Working backwards from your candidate tile to the shape it would require is faster than learning the pattern as a list of pairs.

Actually Safe: Visible and Dead Tiles

Suji tells you a tile is probably safer. There is a category of tile that is genuinely safe, and learning to find them is the core defensive skill.

A tile is safe against a specific opponent if it appears in their own discard pile. They cannot be waiting on a tile they have thrown away — in most rulesets they cannot win on it at all, and in all rulesets they are unlikely to want it. A tile visible in the discards of the player you are worried about is the safest discard available to you.

The next tier is a tile that all four copies of are visible on the table. This is usually called a dead tile. Not only is it unlikely to be wanted, but no player can form a new set with it, so it cannot improve anyone’s hand either.

After those two, the third tier is a tile whose copies are mostly visible. Three visible out of four leaves one copy in play, and the odds any specific player holds it and is waiting on it are low. This tier is not safe in the way the first two are, but it is far better than throwing a fresh middle tile.

The practical procedure, when you must discard and are worried: look at the threatening player’s discards first and take any tile that appears there. If none, look for a fully visible tile. If none, look for a tile with three copies visible. Only when all three tiers are empty do you fall back on judgement and suji.

This procedure is mechanical, which is its virtue. Under time pressure, a beginner’s instinct is to throw whatever looks least useful to them, which is close to the opposite of what defence requires.

There is a fourth tier worth knowing for completeness, though it applies less often. A tile that an opponent has just discarded and that has since been passed over by everyone is unlikely to be a winning tile for anyone, simply because three players had the chance to claim it and none did. This is weaker evidence than a discard pile appearance, but it is a useful tiebreaker when your first three tiers are empty.

一万 · 1 Crak四万 · 4 Crak七万 · 7 Crak中 · Red Dragon中 · Red Dragon

When to Abandon Your Hand

Defence becomes a real decision only when it costs you something. If your hand is hopeless, folding is free. The hard case is a hand that is close and a threat that is real.

Three conditions together justify folding. Your hand is two or more shanten, so you need more than one lucky draw. The game is late, so few draws remain. And an opponent has shown clear strength by calling tiles and discarding from a narrow range. When all three hold, folding costs little because the hand was unlikely to complete anyway.

The mirror image applies to pushing. Push when your hand is at tenpai, because a single tile finishes you and your wait may land before the threat resolves. Push when your hand has real value, since abandoning it wastes more than the risk of paying. Push when the threatening player committed early, which usually means a cheap hand rather than a dangerous one.

There is a middle state that causes the most errors: one tile from ready, with a live threat and several draws left. Here the decision turns on the size of your wait and the size of the threat. A wide wait with three tile types still live is worth defending. A single-tile wait that you can see is half dead is not.

The pattern to internalise is fold early, push late, and never abandon a finished hand out of nerves. Folding a hopeless hand in turn five costs almost nothing. Folding a tenpai hand in the final draws throws away everything you built. Beginners do the reverse, staying in with nothing and then bailing once the hand finally becomes live.

It helps to be precise about what folding costs, because the cost is smaller than it feels. Folding means you stop adding tiles to a hand that was unlikely to finish. You do not forfeit anything you already held, and you keep the option of returning to attack if the threat resolves or a safe opportunity appears. Folding is a pause, not a surrender, and treating it as reversible makes the decision much easier to take.

Defence in Practice

Put the mechanics into a sequence you can actually run at the table, because defence fails under time pressure if it requires judgement calls on every tile.

First, track one opponent. The player who has called tiles and is discarding narrowly. Ignore the others until they show similar commitment; splitting attention four ways means you track nobody well.

Second, maintain a mental list of safe tiles against that player. Their discard pile is the list. Any tile you hold that appears there is a free discard you can take whenever you need one.

Third, when you must choose between continuing your hand and discarding safely, ask how far from ready you are. Engine evaluation of your own hand gives a number rather than a feeling, and two or more shanten late in the game is a clear signal to stop building.

Fourth, count visible copies of the tiles you would need. If your remaining path depends on a tile whose copies are mostly gone, the path is not real, and that alone justifies folding regardless of how close the hand looks.

Fifth, accept the result. Defence is measured across many hands, not one. A session where you deal in once instead of four times is a success even if you never win, because the scoring asymmetry means not paying is worth nearly as much as winning.

The satisfying part of defence is that it improves immediately. Attacking well requires reading three opponents and building a hand at the same time. Defending well requires tracking one player and knowing three tiers of safe tile. The second skill is learnable in an afternoon, and it will show up in your results the same week.