What professional mahjong taught me about decisions

All-last, the final hand of the game. You’re in second place, 1,500 points behind the leader, and you’re not the dealer. You reach tenpai (one tile away from a complete hand) with something cheap: all simples, all sequences, closed, worth 2,000 points, waiting on either side of a 5-6, so any 4 or 7 of that suit wins. Do you declare riichi?

The reflex says yes, because riichi adds value. Here it’s the wrong call. Any win at all, even 2,000 off a third player, puts you in first and ends the game, so the extra value buys nothing. What riichi does cost you is a 1,000-point deposit, a hand you can no longer change, and a table that now knows you’re ready. The leader wants the game to end without you winning, and your riichi tells them exactly when to stop pushing and fold. Stay quiet (damaten) and they keep going for their own win, which is how a 4 or a 7 ends up in their discards.

Put the same tiles in East 1, the first hand, with everyone on even scores, and riichi is the standard play. It nearly doubles the hand (3,900 instead of 2,000 on a ron, more with a self-draw or ura-dora, the bonus tiles that only count for riichi), and it forces three opponents to decide whether to fold. Same hand, opposite answer, because the goal changed.

I’ve been a pro with the Japan Professional Mahjong Association since 2017, and I work in customer success in Japan, so these are the lessons from the table that I use at work. (How I got to the pro tables is a separate story.)

Riichi or damaten in a renewal

A renewal at current terms that already meets your goal is that all-last hand: if it closes, you win. Pushing an expansion into the same conversation is declaring riichi on a hand that already wins. The number looks bigger on paper, but you’ve shown your hand, procurement gets careful, and you’re committed to a position that’s hard to walk back. Close the renewal quietly and bring the expansion later as its own hand.

Early in a relationship it’s East 1: real usage behind you (a good wait) and nothing at risk yet (a flat score). Make the bigger proposal there, and make it clearly.

Push or fold is risk management with half the information

When an opponent declares riichi, you know two things: their hand is ready, and which tiles they’ve thrown. You don’t know what they’re waiting on. Every tile you discard from then on is a small bet on incomplete information: if it’s their winning tile, you deal in and pay for the whole hand yourself. Japanese players call the skill oshi-hiki (押し引き), push and pull.

The inputs are few. How close is your hand? What is it worth? How good is your wait? How dangerous is the tile you’d have to throw? A ready hand worth a mangan (8,000 points) on a two-sided wait pushes almost anything. A cheap hand still one tile from ready, facing a dangerous discard, folds: you break it up and throw only safe tiles, starting with genbutsu, the tiles the riichi player discarded themselves and therefore can’t win on.

What doesn’t go into the decision is how long you spent building the hand; those turns are gone either way. Avoid the half-push too. Throwing two dangerous tiles and folding on the third takes the risk and gives up the reward. Decide early, then commit.

Deals work the same way. How close is it to a signature? What is it worth? Can it close several ways, or does everything hang on one person saying yes (a single-tile wait)? Against that, set what pushing costs: a discount that becomes next year’s floor, or weeks of the team’s time. When the customer has clearly picked someone else and your hand is cheap and slow, fold, and keep the relationship for the next deal.

Judge the decision, not the result

The way I put it in Japanese is 切るべき牌は切る: you cut the tile you should cut, even when it might deal in. Say you push a ready mangan into a riichi and your tile hits a 12,000-point haneman. If the push was right, it’s still right. Expected value is what you’d average if you faced the same spot a thousand times, and one result tells you very little about it.

Mahjong gives you thousands of hands to learn this. Work gives you far fewer decisions, so a single outcome teaches the wrong lesson even more easily. Write the call down before the result comes in: what you knew, what you chose, and what would have changed your mind. Then review against that note, not against the outcome.

Read the table, read the room

Discards are public, and they tell a story. A player who hasn’t thrown a single bamboo tile in ten turns while discarding everything else is probably collecting bamboo for a flush. A player who has been throwing each tile as they draw it (tsumogiri) and suddenly throws one from their hand (tedashi) just changed something. None of this is certain. It moves the odds, which is all you need.

Stakeholders leave discards too. The buyer who asks about everything except price probably doesn’t set the price. A weekly call that quietly turns monthly, or a new name on the thread from a department you’ve never met, is a tedashi: something in their hand changed, and you want to know what before you push. In Japanese companies the decision is often shaped before the meeting, through nemawashi (quiet consensus-building) and ringi (the approval document that circulates for sign-off), so what you see in the room is a small part of the table.

The hand after a big deal-in

You deal into a haneman in South 1 and drop from first to last. On the next hand, everything looks pushable. Now, don’t get me wrong: being further behind can justify a bit more aggression in a close spot. Being angry never does. The honest check is one question: would I push this tile if I hadn’t just lost 12,000 points?

At work it’s the escalation you send because you’re annoyed rather than because it helps, or the next deal you chase harder because the last one fell through. Those points don’t come back from the next hand. Slow down for one hand, then play normally.

Online and offline are different games

I came up on Tenhou, and I play at real tables in Japan now. They’re different games, and neither is the better one. Online is fast and anonymous, you see a huge number of hands, and at the higher ranks Tenhou punishes fourth place much harder than it rewards first, so strong online players get very good at avoiding last. At a real table the rules change from venue to venue, the other three players remember how you played last time, and everyone can see you hesitate before a discard.

Work splits the same way. A Slack thread across time zones is online: fast, written, easy to forward, read by people you’ll never meet. A meeting room in Tokyo is offline: slower, in Japanese, and the silence after a proposal says as much as anything spoken.

Don’t play on autopilot

Big bad pushes are easy to spot in a replay. What costs more over a season are the small discards you make without thinking because they’re usually right: keeping the wrong tile, or throwing one out of habit.

Renewals and onboarding calls are the everyday discards of customer success. Run them from last year’s template and most go fine, which is exactly the problem: the one that needed a different play looks like all the others until it doesn’t.

Don’t play on autopilot. Think before you discard.