Yes, now in hindsight I agree with both of those statements. I was focusing too much on finding the “correct” move and I forgot that go is a game of constantly adapting to the situation and to the opponent’s moves. Especially about the second statement, sometimes it’s hard to fight against the feeling that you might need to sacrifice a few stones to achieve a better result, but that’s the reality of the game. Thank you for this whole conversation, I learned a lot!
3 Likes
Yep, thank you! I noticed the mistake: be careful when changing the direction of a ladder haha! I edited my post with a better solution 
One trick I learned recently is if you have stones on the corner and side star points then there’s no ladder that squeezes through the gap: it will hit one of them. This is based on the gap being 5 lines and a ladder path being 6 lines wide as you marked with crosses and triangles.
But I don’t think I’ve ever used that in a game. I can either tell at a glance if it’s a single stone in the middle of the path, or I read out the ladder if near edge of path or some more complex position it runs into.
3 Likes
Uberdude’s point about star points is one of the advantages of a handicap ⩾8 stones!
If you have trouble reading ladders, then I think it is important to practise, and not to duck the challenge if one arises in your game. But there are some tricks that can make it easier:
- You do not have to imagine every intervening stone; if you find it easier, just follow one of the four diagonal lines to the vicinity of the “obstruction” (potential ladder-breaker), then work out carefully where the other stones would be and continue from there.
- If the “obstruction” is a single stone or a simple position, and is not too close to the edge(s), instead of reading all the way across the board, imagine shifting the obstruction diagonally till it is close to the ladder. This is often easier mentally, and gives the same answer.
- (Thanks to Guo Juan’s Internet Go School for that one!)
- Even if it is at the edge(s), you can try shifting them along with the rest of the position.
- If you lose track of the zig-zag in the middle of the board, the star points can help you stay on course.
One unimportant thing about ladder problems is that they are computationally a very hard class of problem (PSPACE-complete). This has no practical relevance to ladders in normal games or even in tricky puzzles.
Most of the time all you have to consider is the only atari that only allows your opponent to get back to 2 liberties. Just occasionally it gets tricky, either because you have a choice (as you saw close to the edge) or because other stones mean that the ladder deviates from the simple zig-zag. This is more relevant in puzzles (which get way harder than the Valentine puzzle above!), or if you insist on playing ladder breakers and makers and fighting in the path of ladders.
If you do chase a group right to the edge, you get, as mentioned, a choice once it reaches the second line. In this case, you should not thoughtlessly continue the zig-zag, as one of the other choices may leave your position a little better.
3 Likes
Misreading ladder is very common, even at high kyu level. Most of the time, players misread, not because the situation is tricky, but because they didn’t care to read precisely.
2 Likes
Do you mean they do not care enough (feel strongly / think it matters enough), or do not take enough care? Why do you think that is? You say it is very common, do you think it is more common than other reading errors of comparable difficulty?
Perhaps they are too ready to rely on unreliable heuristics, e.g. a stone on one of the 6 diagonals breaks the ladder, or the closest stone is all that matters.
I think they don’t take enough care to visualize stones in a precise way so they don’t see that at some point a stone is in atari, or rely on unreliable heuristics (i.e. they are aware that the heuristics are unreliable but are too lazy to check).
As for the question “is [it] more common than other reading errors of comparable difficulty?”, let’s say that 2-3 kyu players may misread a 15kyu-level ladder every 50 games (I put a random number here, I didn’t make statistics). I don’t know what problems of comparable difficulty are. Ladders are usually easy to read, in the sense that they don’t require high level skills, but may require some effort.
To illustrate carelessness: this game opponent vs. jlt was played on KGS one year ago. I was 1k, my opponent was about the same level. The game was played with long time settings (45 min main time + byo-yomi).
1 Like
Ladders are easy in the sense that the next move is easy to find (in most situations), but hard in the sense that one has to visualise a large number of moves. Perhaps they are also off-putting to some people because it always takes a certain time to work them out, unlike a tesuji problem which you may see in a flash — or struggle with for ages! I can imagine that both these distinctions mean that people can be good a reading one sort and bad at the other, while it is the other way around for others. I guess a lot depends on your time limits as well — and I prefer games where I have time to think deeply now and then.
1 Like
That does surprise me. I can understand the initial (careless) mistake, but I would have thought that once there were a couple of stones in the ladder they would have read it properly. Perhaps in the end they felt the final atari would be a good place to resign?
P.S. The score graph looks like a cross-section of the ocean floor!
He spent 15 seconds before playing the second stone in the ladder (H11), then 42 seconds before playing H10 and 43 seconds before playing J10. He probably made the same reading mistake over and over again. After that he assumed I was the one who misread and played subsequent moves in less than 1 second.
1 Like
I suppose by the time he played O4 he had stopped thinking at all. I suppose it all says something about psychology.
Mind you, one hears people say that you can make it to 1 dan on shape alone with no real reading — maybe he was such a one.
Once you’ve played out several stones of a ladder you are captured in, you might as well keep going because:
- You’ve lost anyway
- It’s a game of chicken, maybe opponent nerves break and they abandon it, I’ve seen an example like this on the forums here
- Maybe opponent was wrong and you aren’t actually captured
- You’d like to see the end
- Good place to resign
- It’s funny
There was a ladder played out on the top boards of the London open a few years ago, between two 4ds, quite a crowd gathered for a chuckle.
6 Likes
I found some interesting OGS games.
-
Move #40
In their chat, she said she didn’t wear glasses, but I wonder if she was drunk.
( •͈ᴗ⁃͈)ᓂ- - -♡ -
Move #105
Again, you can read that white could not pull out the four stones right?
友谊赛 -
This one was mine. It was not that I forgot to read the ladder. I was simply blind.
Glunkolin vs. 食全酒美
3 Likes
First one there seems more like the mistake a 9d bot with too few playouts would make, rather than mistake a real human 9d would make. Second one is possible to be a human mistake, but again unlikely for a 9d human but very likely for 9d bot on low playouts.
1 Like
That one’s worth remembering. There are many joseki-induced ladders that will be broken by a hoshi in the opposing corner, but not by a sansan in the opposite corner.
I think the reading times can be explained easily.
Reading a ladder is relatively easy. Hence why they only spent 15 seconds on the first ladder move.
But then if you play a ladder move too, they immediately know that one of you two has misread the ladder. And, if the ladder is played out, then it’s effectively throwing the game away. So there is a lot at stake, and we know that one of you is reading it wrong. Thus once you’ve both played a ladder move, it makes sense to spend a lot of time re-reading the ladder, before playing the second and third moves. That explains the 45 seconds before H10 and J10.
And then, once several stones have been invested in the ladder, it’s a bit late to suddenly decide to give the stones up and stop playing it. You might as well play it out in its entirety, and then whoever was wrong will resign.
1 Like
Unless the bot is trying to replicate AlphaGoZero, I think the playouts are a very bad place for a bot to read ladders. Ladders and extremely easy to read. You could program a ladder-reading algorithm pretty easily with almost no knowledge about fancy AI algorithms.
The output of this ladder-reading algorithm can then be used as input for a smarter AI.
Relying on the playouts to read the ladder seems pretty inefficient. Ladder sequences are incredibly long, compared to almost every other important sequence in a go game, but they are also extremely linear, with only one possible move at every step and no branching.
The first AlphaGo paper in Nature showed the “feature planes” used as input to the neural net. The neural net sees a preprocessed version of the game board. You can think of this as if the neural net was a go player that looked at the board; but before they do, another player scribbled on the board with a lot of small information about every intersection. For instance, on every stone, someone scribbled how many liberties that stone has; and on every empty intersection, someone scribbled whether playing here would be a successful ladder-capturing move or not, and whether playing here would be a successful ladder-escaping move or not. All of these scribbles, counting liberties, reading out ladders, etc, are computed using relatively trivial algorithms.
AlphaGoZero and AlphaZero got rid of all those preprocessing steps. Not because it made a better go-playing AI, but just to prove a point about how it was possible to build a game-playing AI without providing human knowledge about that game. KataGo is a pretty convincing evidence that these liberty-counting and ladder-reading preprocessing steps are important and that removing them is not a good idea if your goal is to implement an efficient go-playing AI.
5 Likes
That’s one of my recent games here. My 5k opponent and 3k me both spent some time on the first couple moves of the ladder and then played it out quickly until the crucial second line move, then my opponent resigned. I think this time pattern happens a lot with misread ladders.
1 Like