Tile Theory at Two-Shanten (2)
Next comes the pattern with one completed set and four taatsu blocks.
When there are no obvious floating tiles, the choice is basically between two plans:
(1) Drop the weakest taatsu.
(2) Break a three-tile taatsu block back into a single taatsu and keep all four taatsu blocks.
Recent research tends to favor line (1), so this site also treats (1) as the default approach.
Now let us look at actual hands.
Example 1
Tsumo 
=> Cut
The souzu taatsu is a nido-uke shape against 
, so this block should obviously be the one you remove.
Cutting is a bad move because it reduces the acceptance toward one-shanten.
Example 2
Tsumo
=> Cut
The kanchan wait on is 4 tiles, and the shanpon on
is also 4 tiles.
If the numbers are equal, would cutting and making Pinfu more easily be better?
Some older strategy books do indeed give cut as the answer.
But the real difference appears after the hand reaches one-shanten.
Change A
Tsumo 
Change B
Tsumo
It is obvious that the lower one is the better one-shanten shape.
still works as an effective tile even after the hand reaches one-shanten.
By contrast, are basically only waiting to be discarded.
Also, once you drop the kanchan, the hand becomes more flexible.
You can keep one safe tile in reserve, and if you later draw
,
the acceptance toward one-shanten widens even more.
Change C
Tsumo
Example 3
Tsumo
=> Cut
If you think in terms of “drop the weakest taatsu,” the pinzu block is clearly worth keeping, because together with the shanpon it has 8 tiles of acceptance.
So the comparison is between the manzu and souzu kanchan blocks.
The souzu side changes into a ryanmen more easily.
On the manzu side, even drawing only creates another nido-uke shape, so the correct play is to drop the manzu block by cutting .
Example 4
Tsumo 
=> Cut
If you understand the rule “drop the weakest taatsu,” you should not hesitate here either.
You compare pair blocks in the same way: by the kinds of good-shape improvements they allow.
For example, the pair of can become a three-sided wait on a draw of , and
are all still effective tiles.
By contrast, the isolated pair
has only two direct acceptance tiles, making it the weakest taatsu block.
So the correct answer is to cut .
And even if you later draw , cutting is still the best line.
Still, there are cases in real play where comparing taatsu is not easy.
Example 5
Tsumo
The comparison between the manzu and pinzu kanchan blocks is difficult here.
Either cut can be justified.
So in this case, cutting and postponing the decision is also a strong option.
In fact, simulation shows that directly dropping a kanchan is slightly better for raw tenpai speed.
However, if later on either or becomes thin, you can still adjust according to the live situation.
Also, if the hand later ends up waiting on
,
the early discard of also makes slightly easier to ron.
So cutting here is not bad at all.
Theory
Original Japanese page: http://beginners.biz/pairi/pairi11.html