Hand Selection 101: How to Pick Your Hand from 70+ Options

Here's the answer up front: you don't pick a hand. Not at first. You pick a NEIGHBORHOOD of hands (two or three related possibilities from the same corner of the card) and you let the Charleston and your first few draws tell you which one you actually live in. Players who slam their finger down on one hand at the deal are playing a lottery ticket. Players who hold a neighborhood are playing mahjong.

Hand selection is the single most important skill in this game. Not jokers, not defense, not the Charleston. Those all matter, but they're all downstream of this one. Deal me thirteen random tiles and the roughly 70 hands on the National Mah Jongg League card (as of the 2026 card, organized into categories like 2468, Consecutive Run, 13579, Winds-Dragons, 369, and the rest), and the whole game is the conversation between those two things. Let me teach you how I have that conversation, because I have it out loud in every one of my lessons.

Step one: sort into families, not hands

When your deal lands, do not start hunting the card for a hand that matches. Thirteen tiles against seventy hands is a haystack problem, and you will get lost in it. Instead, flip the problem: sort your TILES into the card's own families and see which families light up.

My sorting pass takes about thirty seconds and asks four questions:

• Evens or odds? Count your 2-4-6-8 tiles, then your 1-3-5-7-9 tiles. A lopsided count points you at 2468 or 13579 (or 369, which lives among the odds-and-multiples crowd).
• Any runs? Look for three or four consecutive numbers with support: a 4, two 5s, and a 6 is the spine of a Consecutive Run hand in any suit.
• Winds and dragons? Two or three winds plus a dragon or two means the Winds-Dragons section deserves a look. One lonely wind means it probably doesn't.
• Anything weird and wonderful? Three of a kind already? Quints just got interesting. Same number in two suits? Any Like Numbers is waving at you. A soap and some ones and twos? Check the year hands.

Notice what this does: instead of scanning seventy hands, you're scanning maybe two or three SECTIONS of the card, guided by what you actually hold. The card stops being a menu and becomes a map.

Step two: count distance-to-complete, honestly

For each family that lit up, find the one or two specific hands you're closest to, and count how many tiles you still need. This is where beginners lie to themselves, so let me give you the honest accounting rules:

• A tile you hold that's IN the target hand counts. A tile that's "sort of related" does not. Your 8 dot is worth nothing to a 13579 hand no matter how much you like it.
• Jokers count toward any pung, kong, or quint, but never toward pairs or singles. If your target hand needs three pairs and you're counting a joker toward one of them, you're not eight tiles away, you're nine and confused. (More in my piece on joker rules.)
• Count the pairs problem separately. Pairs can't be filled by jokers and can't be called from discards except for mahjong itself, so a hand needing two unstarted pairs is FARTHER away than the raw tile count suggests. I mentally add a tile of distance for every pair I haven't started.
• Check the X or C. A concealed hand you can't call for is effectively farther away than an exposed hand at the same count.

After honest counting, you'll have numbers like: "Consecutive Run, six away. 2468, seven away but two dead pairs. 13579, eight away." Now you know something real.

Step three: carry two or three candidates through the Charleston

Here's the discipline: do NOT commit yet. Keep your best two or three candidates alive (ideally from overlapping families, so the same tiles serve multiple masters) and run the Charleston as a referendum. Pass from the families that didn't light up. Keep flexible tiles (flowers especially, which appear all over the 2026 card) that serve several candidates at once.

Overlap is the secret. A rack holding 4s, 5s, and 6s in two suits is simultaneously drafting a Consecutive Run hand, keeping a toe in 2468 (those 4s and 6s), and one lucky draw from Any Like Numbers. Every tile you keep should ideally serve at least two candidates. When a tile serves only one candidate, that candidate had better be your front-runner.

Then listen to what comes back. Three Charleston passes plus a courtesy is up to a dozen new tiles flowing through your rack. If the evens keep arriving and the run stalls, the table is voting, and the table's vote counts. I've watched students cling to the hand they picked at the deal while the Charleston screamed a different answer at them. Don't argue with free tiles.

Flexibility beats point-chasing

Every new player falls in love with a big hand once. Usually it's a quint hand or a shiny 40- or 50-pointer, and usually it goes like this: they need two jokers and a miracle, they hold a dead rack for the whole game passing up makeable 25-point hands, and they finish with nothing while someone else's humble consecutive run gets paid.

Hear me on this: a 25-point hand that WINS beats a 50-point hand that doesn't, every single time, and it's not close. The points on the card measure difficulty, not virtue. The high-value hands (quints, Singles and Pairs at the bottom of the card) are priced high precisely because you'll usually fail. Chase them when your deal genuinely delivers the raw material (three-of-a-kind toward a quint, a naturally pair-heavy deal for Singles and Pairs), not because the number is pretty. Most games, the winning move is the boring move. The wins are how you keep score; the style points are how you stay broke.

When to commit: the second or third pick

So when does the neighborhood become an address? My rule: by your second or third pick from the wall after play begins, you should be down to one hand (two only if they still share almost every tile).

Why then? Three reasons. First, by that point you've seen the Charleston's full verdict plus a few live draws, and more information isn't coming free anymore. Every draw you spend "staying flexible" is a draw your opponents spend building. Second, discards start mattering. You can't safely feed the table while you're straddling three hands, because everything in your rack is theoretically useful and nothing is clearly trash. Third, and this is the big one: you cannot call a discard until you know what you're building. That 6 crak hits the table, you have two seconds to decide, and "well, it fits two of my maybes" is how you either miss a call you needed or make an exposure that commits you to the wrong hand by accident.

Committing means: pick the hand, identify your true discards, and start playing with intent. It does not mean never switching. If the wall turns hostile and a sister hand in the same family comes into focus, switch. But switch ON PURPOSE, to a specific hand, with an honest new distance count. Drifting between hands isn't flexibility; it's indecision wearing flexibility's coat.

The beginner trap: tunnel vision

The opposite failure of drifting is tunnel vision, and it's more expensive. It sounds like this: "I picked my hand during the Charleston, and I'm going down with the ship." The player needs a fourth copy of a tile with three already on the table. The player needs to pair a wind that hasn't been seen in twenty draws. The player COULD shift one column over on the card and be three tiles closer, and never looks.

Build a habit: every few turns, re-ask the question. What's my honest distance now? Is anything I need already visible three times on the table? (Four of each tile exist, so count what you can see.) Is there a neighboring hand on the card that my rack now fits better? The card doesn't lock when the Charleston ends. The players I'd call genuinely good re-run this little audit constantly, and it costs them ten seconds a turn.

A worked example, start to finish

Let's do one in prose. Say I'm dealt: 2 dot, 2 dot, 4 dot, 6 crak, 6 crak, 8 bam, 3 bam, 5 bam, 6 bam, north, north, flower, joker.

Sorting pass: evens are loud: two 2 dots, a 4 dot, two 6 craks, an 8 bam. That's six even tiles across three suits, with a pair AND a second pair already made. The bams whisper about a run (3-5-6 with a gap at 4). Two norths is a maybe. The odds are nearly empty, so 13579 and 369 are out. My neighborhood is the 2468 section, with a Consecutive Run hand in bams as the backup and the norths as a lottery ticket I'll abandon first.

Charleston: I pass odd tiles and stray junk, keeping every even, the bam run, the flower, and obviously the joker (which never gets passed). Suppose the passes bring me a 4 bam and another 8 bam. Now the run is 3-4-5-6. But look closer: the evens got better too, because that second 8 bam pairs up. Three pairs of evens plus a joker. The 2468 hands that want pungs and kongs of 2s, 4s, 6s, and 8s are suddenly maybe five tiles away, honest count, and jokers can fill any of those pungs. The run hand is also about five away but needs its pairs filled naturally. Advantage: evens.

Play begins. My first few draws bring another 4 dot. That's the referendum: I commit to a specific 2468 hand using dots and craks, my bam run tiles become my discard supply (fed out carefully, low-value first), and every 2, 4, 6, and 8 that hits the table now gets two seconds of my full attention for a call. Notice the norths died without ceremony three decisions ago. That's hand selection working: candidates enter, evidence accumulates, one hand survives.

Practice this one skill

If you want to get better fast, here's a drill that costs nothing: deal yourself thirteen tiles, no opponents, and just do the sort, the count, and the pick, out loud. Ten deals like that teaches you more card fluency than a month of casual play, because you get thirty candidate evaluations instead of three. It's also exactly what we do across the table in private lessons, where I can catch the dishonest counts in real time. And if a term in this piece snagged you (pung, kong, soap), the glossary has your back, or just contact me and ask. Picking the hand is the game. Everything else is delivery.