Guide - Defense Techiques: Sotogawa and Matagi Suji

In this post, we'll look at sotogawa tiles and matagi suji, including checking Houou replays for some related statistics to back up the theory.

Update: The more thorough analysis alluded to in this post has been completed. You can see it in this post. Feel free to reference those charts as you read through here. Update 2: I've done a specific deep-dive into sotogawa, which you can find in this post.

First, let's talk about sotogawa. There's not really a good English term for it, but it refers to tiles outside of those discarded early by a player being safer against that player. For example, an early 3m would indicate that 1m and 2m are safer against them. Usually this is only used with discards that are twos or threes.

As for why, think about what shapes you would cut a three from that could accept a two. There's 334, 223, or 133. Those are both pretty strong shapes, and you wouldn't normally cut the three from them early. For a shape that could accept a one, there's 233, or 113. Again, you wouldn't normally cut the three early.

That's a fairly intuitive explanation. So, do the statistics back it up? For this article, I looked at every winning riichi hand, the tile it won on, and then checked if they discarded an early 3m. Early is defined as the first three discards, and only hands that called riichi on turn four or later will be checked.

Note that I'm only checking what tile they won on, not whether they could win on that tile. But, it will be a good enough approximation. Anyway, here's the percentage of wins that happened on the first six character tiles, when the 3m was not present in the early or late discards.

Now, let's see the table for when there is a 3m present in the first three discards.

As you can see, the percentage for 1m and 2m went down quite a bit. They're not perfectly safe, but they're definitely safer. You can also see the 6m rate going down due to regular suji principles, as a 3m in the discards ensures the 6m is at least half suji.

Next, let's talk about matagi suji. The idea of matagi suji relies on similar shapes to the ones we thought about in sotogawa, the 233 and 334 shapes. Since these are strong shapes that you're incentivised to keep around, if the three comes out very late, the waits around those are more dangerous. For 233, that makes 14 more dangerous, and for 334, that makes 25 more dangerous, so a late 3m tedashi would make the 1245 tiles riskier to throw.

This applies to all number tiles, not just threes. A 6p could come out of a 566p or 667p shape, making 4578 more dangerous.

In theory, at least. Let's look at the data for when a player calls riichi on the 3m.

The percentage of a 14m wait goes up by quite a bit with a 3m riichi. However, the 25m wait only goes up slightly. Interesting. I'm not sure why it's so different. Now I want to do a deeper analysis looking at every riichi discard and how the winrates change. But, that's for another day.

Another thing you can see is that the 6m rate didn't go down as much as it did when the 3m was in the first three discards. A 357m shape is pretty good, and cutting the 3m from it would make a 6m wait, an instant suji trap. Riichi suji is less reliable than regular suji.

So, you can use sotogawa to help identify tiles that might be safer, and you can use matagi suji to recognize tiles that might be more dangerous. One other thing you might wonder about, is what if they have both an early and late 3m?

Interesting! Well, this didn't happen very often, so it's not a huge sample size. I'll do a more thorough look at this later, using more than just the 3m. The data for this post, including the win rates for every tile, can be found in this spreadsheet.