The data story
The research behind the free card
We ran the numbers so you can run the table.
We read twenty-one years of American mahjong cards — every one from 2006 to 2026 — and catalogued all 995 printed hands as plain fact: what wins, what it’s worth, whether it has to stay concealed. Every year was cross-checked against multiple sources, because the world has misprints and marketplace fakes and we won’t build on either.
What 86,977 means
A card looks like seventy arbitrary hands. It isn’t. Every line ever printed follows a quiet grammar — group shapes, number relationships, suits standing in as roles — and once you’ve catalogued twenty-one years of it, you can map everything that grammar allows. Inside the traditional American style, the map holds 16,370 possible hands; across twenty-one years the annual cards have printed only a few hundred distinct ones. Open it to the patterns independent publishers and the world’s other traditions already play, and it grows to 86,977. Each one is catalogued and rated for difficulty. We call it the scored design space. Every card we publish is drawn new from it, machine-checked against the full archive so nothing repeats.
We measure difficulty by playing
The honest way to learn how hard a hand is: play it until it tells you. So we did, at computer speed. For every design, the computer deals and draws toward the hand hundreds to thousands of times over, counting how often it comes together and how fast — because a real game ends, and the table won’t wait while you build a masterpiece. Then we check the machine against the real world: published win rates from millions of actual games, and independent difficulty ratings. Where our numbers disagree with reality, ours move.
One finding worth your time: printed point values track difficulty only loosely. Inside a single point class, hands that share a score can differ sharply in how often they actually come together — same points, wildly different odds. You’ve felt it at the table. We are re-measuring the size of that gap under a corrected ruleset before we print a number for it.
What this buys you at the table
Point values that mean something: when the card says a hand is worth more, the data says it’s genuinely harder. Difficulty you can trust: the easy sections are truly easy, and the showpieces earn their keep. And new hands for as long as anyone will play — at seventy hands a card, 86,977 designs outlast every table that will ever play them.
| Measure | Figure |
|---|---|
| Years of annual cards studied | 21 (2006–2026) |
| Printed hand lines catalogued | 995 |
| Independent third-party cards catalogued | 14 |
| Hand designs mapped and difficulty-scored | 86,977 |
| Designs within the traditional American style | 16,370 |
| Simulated deals per design | hundreds to thousands |
Honesty is the whole point
What we don’t claim
So, plainly:
We are independent. Not affiliated with, sponsored by, or endorsed by any mahjong league, association, or card publisher.
Our point values are estimates. Good ones — simulated at scale and checked against real-game statistics — but estimates, and they will keep improving as play data grows.
We never reproduce anyone’s card. Every card we release is a new selection from the design space, machine-checked against the full archive for overlap before it ships.
A simulation is not your weekly game. Four players, table talk, the friend who always calls your tile — real play is richer than any model. We measure what can be measured and label the rest honestly.
Drawn fresh from the design space
Get this season’s card.
Twenty-one years of cards, read and scored. The result costs nothing.