Seven findings from 1,000 simulated mah jongg games

How much does a strong Charleston really matter? Which hands keep showing up among winners? And how often does everyone run out of tiles?

We put four of My Mahjong Coach’s computer players through 1,000 games to explore those questions. The clearest finding: a player with the highest-rated hand after the Charleston won 42.5% of all games. That leaves plenty of room for the draws, exchanges, and decisions that follow.

These were simulated games using our implementation of the 2026 card, standard rules, eight jokers, and no blanks. They give us a way to examine patterns in the coach’s play and identify ideas worth testing at real tables.

The Charleston brought the average rack more than two tiles closer

Across all 4,000 player-racks, the average number of tiles needed to complete the coach’s current target fell from 7.71 at the deal to 5.45 after the Charleston. That is an improvement of about 2.3 tiles per player before regular play began.

Average Hand Potential rose from 40.7 to 58.7. Hand Potential is the app’s measure of how promising a rack is for a target; it is not a percentage chance of winning.

Those averages follow each coach’s selected target at the time, so they include improvements from finding a better direction as well as receiving useful tiles. The Charleston gave players an opportunity to reshape their plans alongside their racks.

A useful practice habit is to compare your hand before and after passing: how much closer are you, which natural tiles still matter, and which backup remains realistic?

The strongest hand still had work to do

The player with the highest Hand Potential at the initial deal went on to win 33.4% of games. When we identified the leader after the Charleston, that figure was 42.5%.

The later assessment lined up with the eventual winner more often. That association supports paying attention to the quality of the rack emerging from the Charleston, although this run did not isolate the effect of passing skill from the tiles received.

Here is how the 1,000 games finished:

Excluding wall games, a strongest player won 45.6% of games with a winner. Ties count as a strongest-hand win if any joint leader wins.

For a player leaving the Charleston with a promising rack, there is reason for confidence. For someone still searching for a direction, there is also reason to keep looking: more than half of the eventual winners had been behind the leader at that point.

Changing targets was part of the Charleston

In 83.8% of player-racks, the coach changed its target at least once during the Charleston. That includes changing the numbers or suits within the same card line, as well as moving to another line or section.

Even with that distinction, the larger changes were common: 50.1% of player-racks switched card sections at least once during passing.

This is a useful reminder when your first attractive hand starts to feel like a commitment. Incoming tiles can make a different plan worth considering. A change of direction can be a response to better information.

These figures do not mean that switching frequently is a goal in itself. They show how often these coaches revised their plans as the racks developed. A useful exercise is to name your leading target and one backup, then check both after each pass.

Joker acquisition had a strong association with winning

Jokers produced another striking pattern. In the 722 games with one clear leader in total joker acquisitions, that player won 442 times, or 61.2%.

That measure counts jokers dealt initially, drawn from the wall, and acquired through exchanges. It counts acquisition events, so it is not necessarily the number of different physical jokers a player held.

The timing matters. This is a measure taken across the completed game, not a prediction available at the deal. Looking only at the 574 games with one clear leader in starting jokers, that player won 41.1%.

The two figures describe different groups and stages of play. Together, they make joker acquisition an interesting area for further study, especially the contribution of exchanges. They do not establish how much an extra joker causes a player’s chance of winning to increase.

Consecutive Run hands appeared most often among winners

Consecutive Run accounted for 381 of the 932 wins, or 40.9%. The leading sections were:

Consecutive Run line 2 was the most frequent individual winning line, with 131 wins, or 14.1% of all wins.

That is worth noticing, but it does not make the line a universal target. The coaches selected it at the deal for 896 of the 4,000 player-racks. A frequently pursued hand has more opportunities to appear among winners, and these results also reflect the coach’s own preferences.

Concealed hands contributed 134 wins, or 14.4%. Singles & Pairs and Quints produced just three wins each. Those small totals invite investigation into when the coaches pursue those hands and how they play them; they are not enough to dismiss either section.

For players, the interesting comparison is between the hands you regularly pursue and the hands you regularly win with. Familiarity can influence both lists.

East won more often than the other seats

East won 273 of the 1,000 games (27.3%), compared with an average of 22.0% for the other three seats.

The remaining 68 games ended without a winner. East’s lead over the other seats’ average was 5.3 percentage points, or about 24% more wins. Among the 932 games with a winner, East took 29.3%, above the 25% share we would expect if all seats were equally likely to win. The size of that difference suggests more than ordinary sampling variation in this run.

East also finished the Charleston with higher average Hand Potential: 60.6, compared with 58.0 across the other seats. East starts with 14 tiles while the others start with 13, which may contribute to the difference. This run did not isolate that effect from turn order or other seat-related behavior.

All four seats used the same coach configuration. This is evidence of an East advantage in our simulator, not proof that human players gain the same advantage. A follow-up using matched deals rotated through the seats would help separate position from the luck of the initial tiles.

Wall games give us something concrete to compare

68 games ended without a winner: 6.8%, or roughly one game in fifteen. In games that did produce a winner, an average of 41.2 tiles remained in the wall.

These are useful reference points for this simulator. They are not a benchmark every real table should expect to match. Different rules, player experience, and decisions about risk may produce a different pattern.

Recording a table’s wall games over several sessions would give us a better comparison than relying on memory alone. If the results differ substantially, that would be a useful starting point for examining how the simulated coaches play.

What to try in your next game

Treat the Charleston as a chance to improve both your rack and your plan. Track your progress toward a target, revisit your backup as tiles arrive, and stay alert to opportunities during play even when the opening felt disappointing.

For your next few sessions, record three simple things: your target after the Charleston, the hand you eventually won with if you won, and whether the game ended in a wall game. You may discover that your strongest results come from a wider part of the card than you expected.

About the simulation

This analysis used 1,000 fixed, reproducible deals with all four seats running the same default full-coach configuration. The first 100 games were included in the larger run and reproduced their earlier results exactly. Every game’s racks, targets, scores, exchanges, and move log were retained.

The results describe this version of My Mahjong Coach. A larger sample reduces random variation within the simulator; it does not establish human-player win rates or remove tendencies shared by the computer players.

Related reading: Choosing a Hand · Charleston Strategy