Welcome to the club!
AIs are way above the level of any human player. We just canât compete.
But we can learn something:
- when AI tenukis ask yourself: was it really necessary to play another move in that place?
- when AI doesnât tenuki ask yourself: why should I have put another stone there?
Keep in mind that deep analysis leads AIs to a more fighting style of playing: they are more able than humans to play the whole board at a time, instead than focusing on a single place. So perhaps, sometimes, as a human, spending one more stone in your sequence can be a safer way to play. I donât dare to call it more âsolidâ because my âsolidityâ canât compete with the light play of the AIs. But against humans it could work!
I donât know if AIs ever played honte moves.
Look at the win-rate chart (points, not %) and look for big steps.
Move 28 was W+5
Move 29 was W+13
You lost 8 points because of that move.
Extending at F3 was a good way to protect the cut at C2 while keeping pressure on white and gaining point on the bottom side.
Move 30: white escapes
Move 31: black tenukis and loses three more points
You played elsewhere, but your stones in the bottom left werenât safe!
The group in the corner is quite surrounded, while the three stones at E7 are cut. That means two weak groups that your opponent can exploit.
The sequence starting at move 37 does harm your three weak stones.
37: you extend your safe group for points on the side
38: white keeps your three stones separated
39: you play for points (few points)
40: white plays for influence and keeps surrounding those three stones
41: you try to connect, but it doesnât work
That still was a move for a handful of points, but your three stones were worth much more than that
42, 43, 44: white atari and keeps your two groups separated
45: thatâs useless, black canât cut and the three suffering stones became four
47: a small move, in the wrong direction, pushing white against your four stones
48: white ignores and keeps surrounding the four stones
49: black tenukis again
Move 36 was W+13, move 48 was W+31.
Black lost 18 points in that sequence.
Itâs interesting that move 48 was W+31 while move 49 was W+27
Apparently black gained 4 points by just leaving those four stones that were too suffering.
If you want to use AI to analyse your games, you must look for these evaluations that AI does at each move.
Usually each move loses a little.
A good move is when you keep the score unchanged or lose less than one point.
Itâs very common in my games (Iâm about 4k on OGS) that each and every move gives a point or two to the opponent. But they do the same to me, so the balance is fair!
A positive score is very unlikely and Iâd say that the +4 points of move 49 are just because of the quick esteem done by the free OGS AI. I expect that a deeper analysis would bring it down to +0, which is a good move.
My general interpretation of the AI score estimation tool is that in Go you canât really âstealâ anything from your opponent: you just can either keep your advantage or give something to your opponent.
So when we gain something itâs just because our opponent gave it to us (and we were ready to take it, that is not negligible!)