Analysis - Suji Effectiveness Based On Riichi Tile

Suji is safer when you have reason to believe your opponent has a good wait. If they have a single wait or a shanpon shape, then suji won't apply to them. There are many tells that indicate that a player might have a good wait, but here, we'll look at a very simple one: the tile they called riichi with.

For this analysis, I looked at every riichi hand in my dataset of 1,246,147 Houou hanchan (East-South) games. I calculated the wait, and if every tile in the wait had another tile one suji interval away, I counted it as having a suji shape.

This means that ryanmen (34), sanmenchan (23456), nobetan (2345), sanmentan (2345678), and shanpon shapes that happen to be on a suji interval (3366) would be counted, among other less common shapes like pure nine gates (1112345678999) or double entotsu (34555m45666p). Against shapes like this, if a tile is suji, it's 100% safe.

Shapes that wouldn't be counted are the obvious tanki (4), kanchan (35), penchan (12), most shanpon shapes (2277), as well as shapes that might include a suji pattern but also an extra tile, such as ryantan (3334), entotsu (2344499), and tatsumaki (3334555). Though these have a suji interval, they also have part of the wait that doesn't obey suji, so suji is not 100% safe.

This data is segmented twice. First, between first riichis and chasing riichis, and then, between tedashi and tsumogiri. Here are the results.

It's common knowledge that a riichi on an honor tile increases the chance of ryanmen. If they were keeping a safe tile in their hand, the rest of their shapes were probably pretty good. It's also easier to justify chasing an opponent's riichi if you have a good wait. Check out the push/fold charts. It's almost always justifiable to chase with a good wait (given a relatively even situation). So, the wait quality goes up for chasing riichis.

For tsumogiri riichis, they may have been waiting to upgrade one of their bad shapes, but gave up after enough time had passed and just called riichi. Similarly, sometimes you're dama with a valuable hand, but once someone riichis, you chase since they can no longer defend. Tsumogiri riichis that are the first riichi tend to have bad waits, while tsumogiri chasing riichis are almost comparable to tedashi initial riichis.

It's a little interesting to me that chasing tedashi riichis on a 1 or 9 have a worse average wait than a 456 riichi, while when they're first to riichi, a 1 or 9 tedashi has a better average wait than a 456 riichi. In any case, interesting data.