AlphaGo - Lee Sedol Game 5 Move 68 Positional Judgement

[RobertJasiek]

The attached SGF makes a positional judgement for AlphaGo - Lee Sedol Game 5 Move 68.

[hyperpape]

The attached SGF makes a positional judgement for AlphaGo - Lee Sedol Game 5 Move 68.

Thank you for this example.

[Knotwilg]

Hi Robert

Thanks for the analysis.

A stone that jumps to the eye is Black's head sticking out at L6. This stone affects White's prospects in the centre and at the bottom.

In the sequences where you reduce the centre by threatening white's central stones it seems to play a role. Other than that it doesn't seem to play a big role in yourn estimations, while I think it will play a major role in the more qualitative PJ's by Japanese pros (speculative statement).

What value do you assign to the strong presence of the stone at L6?

[John Fairbairn]

I believe a Japanese pro would recommend an amateur to look at the position after White 68, and to find a move for Black 69, by looking at things something like this:

1. The Black and White groups in the top left more or less cancel each other out.

2. A play by in the lower left corner and a play in the lower centre are miai and so cancel each other out.

3. Black's "confirmed territory" (their preferred term) on the right is clearly bigger than White's so White will have to take points somewhere in the centre.

4. Black needs to reduce White's prospects in the centre, but he has little access to the centre himself, or support for paratroops there, so for that reason and on the general principle that he must keep away from thickness, his likely next move will have to be a reduction.

5. On general principles again, Black must focus first on the weakest group (White's upper left).

So it seems to me that Black's strategy is obvious and can be deduced instantly and entirely without precise counting (reduce in the upper left).

(Going off at a slight tangent, I don't think that if a pro tried a precise count that he would ever include anything for White at the extreme top or in the lower left. Maybe he should, but that's not is what is seen in their books. They would instead give weight to thickness.)

I also believe that most pros would prefer to be White, and one factor would be that while Black has sente White has the initiative (because of the preponderance of thickness). Another (related) factor would be that there are no really weak groups and so the game, at pro level, can be made rather straightforward by White.

[Cassandra]

Just made a tiny experiment ...


My beloved wife's (Ikebana teacher = comparatively 5 d in Go // Go player = about 20 k) judgement (in about 30 - 40 seconds) of the board after White's move 68:

-- White is better.
-- Black will have to struggle reducing White's position in the upper half of the board. Otherwise, he will have no chance of winning.


It seems to me that AlphaGo has a lot of "understanding" for the flow of lines, for harmony, and for beauty.

[RobertJasiek]

Knotwilg, my made PJ assesses the leader and, as a side effect, provides a hint for White's initial strategy. For an assessment of L6 or Black's initial strategy, some of the dynamic aspects of PJ can be applied. The most important ones for L6 are: its development directions, the degree of its - great - thickness, aji elsewhere on the board for using the influence of L6. Black's strategy must include an analysis of the moyo weaknesses.

John, it should not be a too big surprise that different corner enclosures have similar current territories. For a different study purpose, endgame, one would calculate per move values and counts for closing versus invading the corner, first playing on the lower side, first on the left side etc. At this moment, however, we do not need the tedious endgame values yet. I would not identify a miai of corner and left side because the situation is, as the game sequence has nicely demonstrated, more interactive.