Recap of 2026-08-05

This is a review of the stream on 5th August.
I played Go with Bots on Playgo.gg.
I summarised what I had seen on the way.

The first match. Vs Azuril (rate:3000). I played Black.

This is the figure white jumped to △.
Black took territory and doesn’t have any groups to be attacked.
So this is a bit good for black.

In actual game, black connected bottom with corner not to be divided.
However this didn’t seem to be good.
When white 4 moved, it is hatd for Black to connect right with right-bottom corner.

In actual game, Black attached and processed as figure.
But if White moved ×, Black would have been divided.
Although in actual game White didn’t move at ×.

By the way alternative to 7, it looks to connect.
But White has a move to hold out with White 8, 10 and this seems to be even.

So Black should connect here.
White can cut if White wants.
But it is not good shape.

For example if White moved 2, White could not escape two stones.

If White attached on the opposite point, Black capture a White stone and would not be attacked in spite that White could divide them.

I won later because White made a mistake.

The second match. Vs Misty (rate:3100). I played White.

Apparently every Bot has a background.
It was interesting.

This is the position Black jumped to △ and divided White.
I thought this was good for Black for now.
However AI says this was a bit good for White.
Since right half had been Black’s sphere originally, it was balanced.
I might be pessimistic to have seen the rate of the opponent.

In actual game, White reinforced bottom’s group.
But this was not good.
With Black’s 2, it became difficult for White to move.

White should reinforce right side urgently.
On the bottomside White would not by attacked if Black moved A then White would B, and if Black moved C then White would D.
And when needed White had a move to make an eye with E.

The rest of the game. I thought I would become bad if I moved ordinary, so I moved White 3. However this was not good as well. Around here blunder led to blunder.

If moved without thinking White would move 3.
I avoided 3 because this would be good for Black.
However, if it were reinforced with 7 and it would be even.

After Black 4, it was one way for human. White had two weak groups and lost the initiative completely.

I won later because White made a mistake. I had to put in a bit of extra effort after the situation took a turn for the worse, but my play might have actually been surprisingly sharp.

So I played two games with AI Bots on Playgo.gg.
General AI wins straight forward if we are bad in the early or mid-game. On the other hand Bots on Playgo.gg doesn’t looks like that. We seem to have plenty of chance after mid game if we become bad. The moves of the Bots on Playgo.gg is human-like such like early supervised learning AI which are Ray or Leela.

I don’t know the detail of the recent developments in AI Go but I image they use Human network of KataGo. Bots on Playgo.gg may be good opponents if you want to play against AI that is human-like.

• Go