Review requested, 13x13

I can’t make sense of the AI, it has huge drops all over the place, so I can’t tell if a move was actually bad as a whole or just bad for the particular reading of the AI.
Also, I really don’t get the huge difference in move 72.
Any help is appreciated, I know 13x13 is not the most common to be reviewed.


I would say that at your current beginner level, you’re quite a bit too inexperienced to benefit from the AI review, you should almost entirely ignore it and just ask for help from human players. Like you’re doing in this post!. :slight_smile:

If you’re curious anyways - I would say that the vast majority of things that the AI is complaining about for both player’s moves in this game would be standard and common instinct for many single-digit-kyu players, and completely basic and elementary for dan players. But of course, at ddk, they may seem far from obvious.

Here’s a basic review here:

Main themes are:

  1. Pay attention to how much “base” groups have and whether they are undermined or not - groups with much less than 10, or that are undermined, are under serious threat and need attention. But only if they’re surrounded and don’t have other eyes - obviously if they’re not even surrounded, or if they have friends nearby, or actually have eyes already, it doesn’t apply. (And also understand that this is a rule of thumb and as you improve this rule will quickly become less useful, it’s only a starting point).

  2. Work on tactics. Stop playing as many defensive moves that are unnecessary (but don’t stop playing the ones that are necessary) and pay attention to what cuts and connections work or not.


Thank you very much. :slight_smile:

I usually understand huge drops in AI (like 80% or more, sometimes the mistake is obvious to me, but only on a basic level. I couldn’t even get that in this game :frowning: ).

I don’t really get this though :frowning:

A rule of thumb is that IF a group is surrounded or otherwise has no way to expand or connect to friends, then you drop straight lines down to the edge and count how many points of “base” would be surrounded in the interior (not counting the spaces on the lines themselves). This rule of thumb applies to “loose” groups that haven’t actually made eyes yet or a completed and known living shape.

If it’s around 10 or more, then the group is probably healthy for now. Or, if you already actually have one eye, or very good eye shape, then only half as much base can be okay - 4 to 6 points would be healthy.

If it’s less, then the the group is varying degrees of more or less healthy, and deserves immediate attention (or might already be dead). If it’s less but you have a move that raises it to near 10, then that move is probably urgent, and failing to do it will have your group die.

Similarly, if you do have 10, but the group is “undermined” by a stone already in place to poke out the base (usually a stone closer to the edge than any of the stones in your group so far, positioned to “poke right in under”), so that the opponent in one move can reduce you to a lot less, then the group isn’t stable. You want 10ish points, and in a way that it’s a stable 10 points. (or to not be surrounded in the first place, or to have a way to escape, or to already have eyes/eyeshape, etc).


I see! Is 10ish points also for 19x19 or it’s more there? Or board size doesn’t apply for this?

1 Like

Heh! It’s the same on every board size. That should be intuitively obvious if you think about it a bit, right?

Because: it obviously doesn’t matter how big the rest of the board is when we’re just talking about whether, locally, some group that has been surrounded (i.e. cut off from the rest of the board) is going to live or die.

Again, it’s an extremely crude rule of thumb. You should learn it and apply it now, and then as you become SDK and better, you should forget it and stop using it because it’s not very good.

But notably, this extremely crude rule of thumb (if taking into account undermining stones as well) would have triggered on every one of the situations in this game where both players repeatedly kept ignoring situations where groups were at stake and whoever played first would live/kill.


Yes, it makes sense now, thank you. :slightly_smiling_face:

When you see win rate doing that, usually it means that the game is close, so players are ahead or behind by few points that can be gained or lost with every move (if they aren’t optimal moves).
So, good news: you and your opponent played a game very even from start to finish. :slightly_smiling_face:
These are the most exciting and funny games.

Also, I really don’t get the huge difference in move 72.

Black attack was too deep and could be captured if white answered differently.


If you haven’t already, now that the concept makes sense, you might go back through the game and see how this concept applies to the groups in the game as an example, to help detect when a group is threatened or threatenable, or dying/dead.

I also added a few more comments around the moves 70s and 80s.

Hope this helps!

1 Like

AI is really really good at go :stuck_out_tongue:

if you can look at its recommendation and it makes sense, learn from the logic you can infer from it’s move…Awesome.

If you cannot get it, just move on.

Few points,

  • Again it is really really good, lets not be too broken up if the most skilled go playing entity is playing above our heads.
  • Even if you understand, if you cannot read like the machine, your best move may not be it’s best move
  • 9x9 and 13x13 to a lesser extent will always be more dramatic. Smaller board, fewer points, every move is more important. If you lose some area in 19x19 you make some somewhere else. in 9x9 you are dead. 13 fits awkwardly in between lol. AI for smaller boards certainly seems to take a very dramatic view.
1 Like