Mahjong Statistics vs Japanese Riichi Players - Who's Right?

By Nicolas Pro8 min read

How important are data and statistics in Japanese riichi mahjong? Do professional players rely only on numbers?

There are many statistics that players collect when they play mahjong online. These statistics can be a great way to learn about the game and how it is played, but there's nothing quite like getting advice from more experienced players in person.

If you're not sure which data to trust, this blog post will provide some insight into whether or not we should take stats collected by others at face value or if we should rely on our own instincts and experience instead! Let's debunk some mahjong myths.

What do you mean by Mahjong Statistics here?

I'm talking about probabilistic data that other players have gathered.

There are two main types of data:

1. Probabilistic: This is the pure mathematical approach to calculating the likelihood of an event. For example, there are 34 types of tile, so if you draw one random tile from all 136 tiles in a pit, you have a 1 in 34 chance. Nothing changes that. This is extremely reliable data.
2. Statistical: This is based on what actually happened in data samples from games. The winning rates of certain waits by a certain turn number, the likelihood that someone is tenpai by a certain point in the game, and more. This is useful, but not always reliable.

Thanks to the advent of Online Mahjong, there has been a boom for collecting and sharing data in the Japanese mahjong community. Even in the West, too.

However, how important are they actually?

Should you take the time to learn and memorize all the edge cases or just play and let your intuition build up?

If you’ve ever wondered about the importance of statistics in mahjong, you’re not alone. After all, Japanese mahjong is a game where at any given time, you can play the odds and still lose many games. However, I’d like to clear up some misconceptions about data and how some professional mahjong players, myself included, like to think about the game.

Do Mahjong players use “feeling” or “statistics and data”?

Ah, the old debate. Occult or digital, John or Paul, religion or science, chicken or egg. Maybe not that last one.

Do Japanese mahjong players constantly use data to calculate and decide what the best move is?

Well... yes, but no. I mean, partially!

On the one side, betting all your chips on an event that has zero chance of happening will always be a losing decision. So you cannot and should not ignore the math.

On the other hand, a riichi mahjong aficionado has an advantage over someone who doesn't know the intricacies of Japanese mahjong--they can see through certain tells such as waiting patterns or whether a player is tenpai based off how they play. Once you have been read, there is less of a need to rely on a purely mathematical view.

A riichi mahjong aficionado can see through certain tells such as waiting patterns or whether a player is tenpai based off how they play.

You might be surprised to read many of the strongest players right now lean towards a more “humanistic” approach, taking into account not only data, but a lot more meta-information, adopting bluffs and other strategies in the heat of the moment, much like Poker players have been doing for decades.

Statistics and data pave the path for having a stable, “standard” choice when everything else is equal.

Orthodox digital players might tell you that the right way of playing mahjong is simulating or calculating the Expected Value of every single discard to decide what the correct choice is and that the result of such analysis will always yield "the one holy, correct discard". Don't worry, that is not what real "digital" play is all about.

EV (in a vacuum) is indeed a vital factor for deciding discards, but it’s not the only one. Years of experience will eventually teach you about additional factors that can’t (yet!) be accurately measured in the middle of a round.

Mahjong Statistics and Hand Reading

Hand reading is often overlooked by "1-Player Mahjong" enthusiasts. See:

Say for example:

Do you know the probability for hon itsu tenpai for somebody with 2 calls of the same suit during the second row of discards?

💾

Reading a Hon Itsu hand

Okay, that was an easy one, there's probably a chart out there for that one already.

But, there's a catch now, there are many other factors to consider in addition to this.

What about if she changed her hand 4 times after her last call?

How about if she discarded a dora ryanmen shape early on?

What if she did both of the above?

Wait, there's more: How about if you have seen, and you know, that she’s not the type of player who usually calls twice for less than 3900 points?

Did the chance now get any higher?

What if, while she’s right-handed, she's discarding everything from the right end of her hand? (The likelihood of hon itsu increases in these cases!)

Alright, finally, what if she loses her rhythm or drops the pace before each yakuhai she discards? And, what if she doesn’t? What if that yakuhai she discarded was the 4th one of its type? What if it was a live yakuhai instead? What if she stutters whenever the person to her left drops a tile in that suit?

And let’s not even get started with what tiles in particular she called, didn’t call, or what her discard area might look like. You see where I am going with this...

The point is: there’s a lot of meta-information on the table and although statistics and data can help, they cannot yet take an exact measure that takes all this data overload into account.

The data available would not be enough to cover every specific case, and even if it were, it would be next to impossible to memorize every single case and combination.

Any given mahjong hand comes with much more data than you can hope to process as a player

It's up to you to either use this data to your advantage and draw your own probabilistic conclusions, or to just ignore it and revert to static numbers.

In fact, if you're inexperienced, you're better off relying on what's proven and true. You will miss on some chances of winning, but at least you won't actively increase your chances of losing by making poor reads.

Less experienced players are better off relying on data than on their experience

By the way if you would like to see some more examples of how to read Hon Itsu hands, you can subscribe to my Patreon and get access to special posts like that.

Statistics come from different sources: the source of the data matters!

Most data nowadays comes from Houou (Phoenix) tables on Tenhou.net which is a completely different ruleset (and player base) from your average mahjong tournament or club.

Phoenix Players represent less than 1% of the mahjong playing community. Usually, the people you encounter in real life will not be that good, neither play that strong of a defense, and surely they will not play in the same way if the rules are different.

For example, you are likely to encounter sakigiri (getting rid of semi-useful tiles earlier than you'd like to) and betaori (dropping any chance of winning the hand to defend) more often in Tenhou than in High Rate Tonpuusen parlors.

Please, pick your statistics carefully; ideally use data from the same ruleset that you're playing.

Mahjong has a lot of “gray” areas when it comes to decisions. Strong players are just good at telling apart the whiter shades of gray from the darker ones. -Asapin

How much should I rely on mahjong statistics?

When you have no other relevant information, data and statistics should always come first.

So, yes, you should learn at least some basic statistics like how likely are you to tsumo on a ryanmen riichi, or how often is ippatsu tsumo expected (this will also help you manage expectations and frustration).

You can find all sorts of amazing statistical data in Amber's blog.

However, as you become stronger and you start to pick up more and more reliable info, you can start twisting those numbers around a bit - you can shape the matrix to your liking.

💾

"What are you trying to tell me? That I can dodge expensive hands?"

No, friend. I'm trying to tell you that when you're ready, you won't have to.

In all seriousness, deviating from statistics is not necessarily a mistake, it is more like a privilege.

Deviating from statistics is not a mistake, it is a privilage

You’d be surprised at how legendary mahjong players manage to turn the tides in their favor and beat the odds thanks to their experience. Legendary Meijin Tsuchida Kosho is one such example.

And although he plays a very unorthodox mahjong, you might be surprised he's very knowledgeable about probability and statistics. He just chooses to ignore them once he's onto something!

Learn the rules like a pro so you can break them like an artist. -Picasso

In the end, it is up to you and your personal preference. Do not be afraid of using both data and instinct!

Need more practice? Try these riichi mahjong tools and resources.

If you'd like to tap into my expertise, schedule a mahjong lesson with me and I'll be happy to help you achieve your goals.

Please share and like this post if you found it useful. Good luck at the tables!