Best Mahjong Opening Hands: What to Keep in Your First Discards

Your First Discards Decide the Hand

The opening is the part of mahjong where skill compounds fastest. A single early discard rarely matters, but the pattern of your first four or five discards sets the shape of everything that follows. Get them wrong and you spend the rest of the hand recovering. Get them right and the hand plays itself. See also Mahjong Strategy: How to Actually Win More Hands.

The reason the opening is so consequential is that it is the only point where you have maximum freedom and minimum information. All thirteen tiles are yours and nobody has shown anything. Every later decision is constrained by what you have already thrown and by what your opponents have revealed. The opening is the one moment you are choosing your hand rather than reacting to it. See also Mahjong Tiles and Their Meanings: The Complete Guide.

So opening strategy is really about which shapes deserve to be kept. Some starting tiles are obviously disposable — isolated honours, tiles with no neighbours. Others look similar but differ sharply in how many future tiles they accept. This guide gives you the shanten counts for those shapes so the choice stops being a guess.

A useful frame before the details: a hand is four sets plus a pair, so thirteen tiles must eventually become those five blocks. In the opening you are not trying to complete anything. You are trying to hold five blocks that can each become something, and discarding the tiles that cannot belong to any of them.

There is a second reason the opening deserves separate study, and it is about mistakes rather than upside. Opening errors are the easiest to repeat unconsciously, because each individual discard is small and the feedback arrives many turns later. A player who habitually keeps isolated honours will never see a single turn where that choice visibly costs them. They will only see a general pattern of hands that never quite came together, and that pattern is much harder to diagnose than a single blunder.

One warning that applies throughout. These numbers describe tile layouts in isolation. They ignore what is visible on the table, what your opponents are doing, and what suit you are being handed by the draw. Use them to compare shapes, not as a formula to follow blind.

The Shapes Worth Keeping

Not all two-tile blocks are equal, and the differences are larger than most beginners assume. Given the same rest of the hand, three partial blocks accept different numbers of completing tiles: a two-sided block such as 2-3 accepts four tile types, a closed block such as 1-2 accepts four but from one end only, and a gap block such as 1-3 accepts four into the middle.

The wide-open block is the strongest simple shape. A block like 2-3 accepts 1 and 4, which is two tile types across two copies each, for four tiles’ worth of improvement. It also keeps more of the suit available, because 1 and 4 sit at opposite ends of the block rather than in the middle.

Pairs deserve their own mention because they are more flexible than their two tiles suggest. Run the engine on a hand whose remaining blocks are complete except for a 2-2 pair and it reports four waits; the same hand with a 2-3 block reports three. The pair can finish as a triplet or, if the rest of the hand provides one, serve as the pair the completed hand requires. That dual role is why pairs are worth keeping even when they look passive.

The most valuable shape in the opening is the compound block, and it is the one beginners consistently fail to recognise. Hold 2-2-3-3 in one suit and the engine reports six waits. Hold the equivalent 2-3 and 4-5 and the same hand reports three. The compound shape overlaps: it can develop toward two sequences, toward a triplet plus a sequence, or toward a pair plus a sequence, so it accepts tiles that help in any of those directions.

The practical upshot: when you see overlapping tiles in one suit, treat them as one asset rather than two separate problems. Their value is not the sum of their parts. It is the set of futures they allow.

There is a ranking worth memorising, from weakest to strongest, for the two-tile blocks you will see most often. A closed block pointing at the far end of a suit is weakest, because it accepts from one direction and the tiles it needs are near the edge where fewer shapes live. A gap block is better, since it accepts a specific middle tile from either side. A two-sided block is better still, accepting from both ends. A pair is roughly as flexible as a two-sided block but in a different way, and a compound overlapping shape beats all of them.

None of this means you should keep a weak block over a strong one in every case. The rest of the hand constrains the choice — a suit you already hold three sequences in has less room for a fourth, and a block in a suit you are largely ignoring is worth less than its raw flexibility suggests. What the ranking gives you is a default, and defaults are what you need in the opening when there is no other information to use.

二条 · 2 Bam二条 · 2 Bam三条 · 3 Bam三条 · 3 Bam

The Shapes to Throw Away First

The mirror image of keeping good shapes is identifying dead tiles, and the opening is when they are easiest to spot. A tile is dead when no realistic path leads from it to a set plus a pair.

Isolated honours are the clearest case. A lone East wind has no neighbours, cannot form a sequence, and needs two more identical copies to become a triplet. Two of those copies might sit in other players’ hands or in the wall. Early in the hand it is the most disposable tile on the table, and discarding it first costs nothing.

Isolated terminals are the next to go. A lone 1 or 9 in a suit can only form a sequence one way, and only with two specific partners. Holding that 1 while the 2 and 3 never arrive is a tile you carried for nothing. Late in a session you will notice how often the tiles you finally throw away are the same isolated terminals you held at the start.

The subtle mistake is discarding a middle tile because it looks unconnected. A lone 5 has neighbours on both sides and can join blocks in four different directions. It looks like nothing and behaves like an asset. Beginners throw these and keep their lone honours, which is precisely backwards.

There is a timing rule that captures it: discard in order of shrinking options. Honours have the fewest paths, then terminals, then middle tiles. Following that order means you spend your early discards on the tiles that could never have helped, and reach the middle game still holding the tiles that still can.

One subtlety that catches players who have learned the rule: isolated honours are disposable, but a *pair* of honours is a completely different asset. Two identical winds can become a triplet with a single draw, which makes them a legitimate block, and a triplet of your seat wind or the round wind carries bonus value in many rulesets. The rule is about isolated tiles, not about honour tiles as a category. Discard the single East; keep the two of them.

The same distinction applies to terminals. An isolated 9 is nearly worthless. A pair of 9s is a valid block with a path to a triplet. Beginners who learn the terminal rule too literally will throw the second 9 and lose a real asset along with the dead one.

Counting Copies Changes Everything

There are exactly four of every mahjong tile, and every copy you can see is a copy that cannot complete your hand. This fact is worth using from the first discard, not just in the endgame.

The logic is straightforward. If a shape accepts a particular tile and all four copies are visible, that shape is dead and should be reassessed. If two are visible, the shape still works but at half strength. Two shapes that look equivalent can be far apart once you count.

Applying it early feels unnatural because the opening is when least is visible. But even the opening gives you information. Your own hand holds some copies, and the first few discards from each player are public. By the time you have thrown four tiles yourself, a meaningful fraction of the wall is accounted for.

There is a second-order version of this that separates strong players. Holding a tile that your own discard pattern suggests you do not want is a form of disguise, but holding a tile that is visibly scarce is simply a waste. If two copies of the tile your block needs are already discarded, that block is worth less than the same block in a suit where nothing is visible — and the correction belongs in your opening plan, not just your endgame.

The habit to build is counting the tiles that matter to your two or three live blocks. Not all 34 tile types, just the handful your hand is actually waiting on. That is a small enough number to track in your head while the hand is still developing.

A worked example makes the arithmetic concrete. Suppose your hand’s remaining path runs through the 3 and 6 of one suit, and you have seen two 3s and three 6s across the table. Your effective outs have dropped from eight tiles to three. The block is not dead, but it is running at under half strength, and a competing block in a suit where nothing is visible is now the better bet despite looking no different on your rack.

This is the calculation that separates a player who is reading the table from one who is merely looking at it. The tiles you cannot see are the only ones that can still help you, and their number changes every turn. Recalculating it is cheap; ignoring it means planning around tiles that no longer exist.

What a Strong Opening Looks Like

Put the pieces together and a strong opening has a recognisable profile. Several blocks are live, at least one is a two-sided shape, at least one pair is present or forming, and no tile is being carried without a path to a set.

A concrete example. Start with m2 m3 m4 m5 m6 m7 p2 p3 p4 s2 s3 s5 s5 and the engine places you at shanten 0, waiting on s1 or s4. That is a finished-quality hand from the opening — three sequences, a pair, and a two-sided block. It did not require cleverness. It required holding the right shapes while discarding everything else.

Compare a weaker opening with the same tile count: m1 m3 m5 m7 m9 p2 p4 p6 p8 s1 s3 s5 z1. Every tile is isolated, nothing overlaps, and the engine reports shanten 5. Hands like this are not unlucky. They are what remains when a player keeps tiles that only look useful — alternating numbers, one from each suit — instead of tiles that connect.

The difference between those two hands is not the draw. Both are legal thirteen-tile hands. The difference is that one holds five coherent blocks and the other holds thirteen separate tiles.

So the concrete opening checklist is short. Discard isolated honours and terminals first. Keep pairs and overlapping tiles in one suit. Prefer wide-open blocks to closed ones. Count the visible copies of the tiles your live blocks need. And judge your hand by whether it holds five blocks, not by whether the tiles look tidy.

It is worth noting what the checklist deliberately leaves out. It says nothing about which suit to favour, because that should follow from where your tiles already cluster rather than from preference. It says nothing about speed versus value, because that choice belongs to the middle game once the hand has a shape. And it says nothing about defence, because in the opening there is no threat to defend against.

Keeping the opening checklist narrow is what makes it usable. A player who tries to optimise everything from turn one ends up paralysed, comparing tile values that will not matter until much later. The opening has one job — establish five live blocks — and doing that well is enough.

After that, the opening is over and the middle game begins — where the question shifts from which tiles to keep to which tiles are safe to throw.

二筒 · 2 Dot三筒 · 3 Dot二条 · 2 Bam三条 · 3 Bam五条 · 5 Bam五条 · 5 Bam