Kuinobashi is based around shapes that are a completed run attached to a kanchan. For example, 34568. With this shape, you could call 2 or 5 and discard 8 to upgrade it to a ryanmen wait. That would make the 4-7 tiles the ones you were waiting on. Thus, if we saw someone making a call and discard like that, it would warn us of the potential danger.
There are eight of these shapes. Let's look at each of them, the tiles that would be used to call, the tile that would be discarded, and the ending wait.
You might be able to notice a pattern. The tile they call and the tile they discard are in the same suji interval. This is the main tell for a kuinobashi. Then, the dangerous tiles are in the suji one tile over from the discarded tile.
Does the data back it up? Let's take a look at all of these call->discard patterns and look at how the tile win rates differ from average. We're also only looking at the last call the player made.
The wait indicated as dangerous by kuinobashi is in bold and underlined. As you can see, it's very dangerous.
There is another kuinobashi type, though it's much more minor. If we overlap the kanchan and the completed run by one tile, we'll get a shape like 34557. With this shape, you could call a 4 with the 35, discard the 7, and be waiting on 3-6. Let's look at the possible shapes.
Again, the same pattern of calling a tile then discarding the suji of that tile is present. However, if you notice the Call With column combined with the Discard column, you'll notice it's a ryankan shape, such as 135. So, this can just be a false positive for that. Well, let's check out the data.
The tiles are less likely to be won on than your average hand. However, comparing them to the other tiles around them, they are relatively more dangerous, so it may still be prudent to avoid them if you can help it.
Finally, if we move the kanchan one step away rather than closer, we get shapes like 13567. This is more obviously a ryankan + ryanmen, just like the one we just looked at. If you call 4 and cut 1, you'll have a 5-8 wait. This one is more restrictive, so there are only four patterns.
Let's check what the data says about it.
These ones get more dangerous than the previous. The winrate on the outside tile is higher, because the inner tile has had at least one copy used up by the chii itself, so it will of course be won on less in comparison. These calls are actually all present in the previous chart, too.
Keep an eye out next time some calls chii then discards a suji of the called tile!
... If only it were that easy. You may have noticed there exists a pattern in the first group we looked at that does not obey the suji rule. With the 23457 shape, you can call 14 and discard 7 to wait on 36, but you can also call 3 and discard 7 to wait on 14.
With these, they call a ryanmen from one side, then discard a tile a space away on the other side.
Oddly, while most of the resulting waits get an win rate boost, the 345->7 and 756->3 have one that drops. Not sure why.
For all of these, we've looked at kanchans next to completed runs, but there are also cases that come up with penchans near completed runs. These are notable in that they allow the person calling to switch to tanyao.
With these, they call a kanchan then discard the tile that would have created a penchan with the outer tile they called with. Note that if they had a yaku and a ready hand, they would simply win here.Basically all the tanyao tiles in the suit that weren't involved in the call or discard get increased in likelihood. As this is a call that switches to tanyao, that makes sense. I'd bet the tanyao tiles in other suits also go up in danger a bit from this sort of call, but I don't have that data.
No comments:
Post a Comment