Nordic Go Dojo

A Japanese go adage tells us to always read three moves ahead. This adage is also in use in shōgi and, as I just now found out, in chess as well.

I first learned of the idea a long time ago – probably when I was studying a life-and-death theory book – and quickly dismissed it: isn’t it even better if you can read four, or five, or six moves ahead? Why should you stop at three? Of course it is better if you can read further ahead, but I think the younger me rather missed the point of the adage.

Let me start discussing the topic from a completely different direction – but before that, let us again define a new Japanese go term.

Tedomari, or the last move

Tedomari literally translates to ‘end of moves’, so ‘the last move’ is an apt English version of the term. It can literally mean the last move in a game that is worth points, but it is also used to denote the last move of a certain size category. If, for example, first Black plays a 10-point move, then White plays a 10-point move, and then the biggest move Black can find is a 5-point move, then we can say that White got to play a tedomari. As a general rule, you want to play as many tedomari as possible.

Dia. 1 shows an example of a tedomari. Whoever has the move turn in this game will play the half-point move of ‘a’, and their opponent will be forced to play a dame or pass. If we look at these two moves as an exchange, whoever goes first profits by half a point.

Dia. 2 has two moves remaining, at ‘b’ and ‘c’. If the first player plays ‘b’, then the opponent will play ‘c’ and get the tedomari; and if the first player instead plays ‘c’, the opponent can play ‘b’ as tedomari. Although these two moves cancel each other out, we would say that the second player profits, because the second player has played a valuable move just before this position came up.

Tedomari problem

The above problem is from Rational Endgame. Give it a bit of thought before reading on!

Tedomari problems are interesting because they show that the game is not simply about finding the biggest move at all times. Also, the solution to tedomari problems varies depending on the move turn; in the above problem, too, Black’s and White’s correct first moves are different.

Both ‘a’ and ‘b’ in the problem diagram are one-point moves, so counting the values of the moves does not help us get to the solution. What, then, should we do?

There is no choice but to read the sequences out.

Dia. 3 shows Black’s correct move order. Both black 1 and white 2 are one-point moves, but, after white 2, Black gets to play a tedomari one-point move at ‘c’. Consequently, the game ends in a tie.

If Black instead played at 2, White would play at 1, getting the tedomari; in this case, White would win the game by one point.

Dia. 4 shows White’s correct first move. After white 1, ‘d’ and ‘e’ are both worth one point and are therefore miai; White gets to play the tedomari and wins by two points.

If White instead mistakenly played ‘e’ as his first move, Black would play 1, and White would only win by one point.

Although Black’s and White’s solutions to the problem are different, their logic is the same: ‘I play my move, you play your move, and then I get to play one more move.’ This is the core idea in ‘always read three moves ahead’: when you think of your next move, you shouldn’t content yourself to just picking a move that looks nice, but you also need to predict how the opponent will respond. If you have a good response prepared for that, then you’re good to go – and you can go through the same three-step process again on your next move, ad infinitum (until the end of the game).

Tedomari whole-board problem

The above problem is from an ngd league game. It is White’s turn; where should he play? While it is hard to say that this problem has a ‘correct’ answer, we can employ tedomari thinking here as well.

Dia. 5. In the game White played 1, which is decidedly a big move. In response, however, Black can close her territory on the lower side at 2, after which White realises that there are no particularly big moves left on the board. White has erred: Black got the tedomari.

In principle, when you aim for tedomari you aim to leave two similarly big moves on the board after your move. If your opponent then plays one of them, you get to play the other, and your move ends up as tedomari.

Dia. 6. One way for White to fight for the tedomari is to play the slightly strange-looking ‘parachute trooper attack’ on the fifth line with white 1. It is difficult to explain why this white 1 in itself would be a good move, but the idea is a lot clearer when we think about tedomari. After white 1, it seems like Black has two similarly good moves on the board at ‘a’ and ‘b’, but Black cannot get both of them; White is set to get the tedomari. Although this white 1 is not among the ai’s top choices, KataGo judges it roughly two points better than white 1 in Dia. 5.

As we have seen above, go is not simply a game of two players picking the biggest move in order; there is a whole another layer of depth in controlling the number and size of moves that remain after one’s move. While it can be impossible to read until the end of the game, most of us can learn to read three moves ahead with a little practice.