Counting practice

of course timers were around long before the internet and such. And indeed games can be short without a timer. But having a timer does ofter structure. And where there is structure there is accessability in my opinion.

I personally prefer to play with no time myself, especially in person to person games in the real world. thoguh i have never actually played with a timer IRL so my opinion may be somewhat Jaded.

But i also agree with what @Kosh said.

Also… To be fair… if my brain were a computer it would seriously be lacking RAM (Random Access memory)… so having a quick way to process things is always a great idea for me. My head is so full of useless crap.

I have this problem too and the tip is nice. I’ll try it. Thank you.

2 Likes

I am absolutely unable to count a whole board in one minute.
A friend of mine, 2d IRL, talking about counting, told me about the situation in which you are in byo-yomi and have to count, and he clearly stated that he was counting each piece of the board in between playing moves.
That’s another complication having to choose a move while counting and also remember the count through the evolving situation.
But I’d say that he was taking more time than 60 seconds to count the whole thing.

3 Likes

That’s when you need strong time-suji :wink: you say ko threat, I say extra period for counting haha

4 Likes

Like i said in another reply to this topic… i dont think in my suggestion that i have the times perfectly set out as it were. they were just sort of examples… time adjustment and so on should be changed to better suit the level of player.

for the record… i also cannot count a board accuatly in a minute… I can get a fairly good understanding of who is ahead, and where… but i dont exaclty know the numbers in that time. So yes i do agree that if this were to be made, that its set up should most certainly contain diffeent ones than the ones i used as examples.

2 Likes

I hear you. :smiley:
That’s weird: by the first 20 minutes of that game I was sure he was far ahead. The territory on the right was so big.
But it was just 10 points difference.
So I’d say that counting would have been useful in order to know that a 10 points capture would waste the game! :wink:

But he also managed to capture three stones, so I don’t really get where all that difference came from. Weak endgame?
I’d like to see AI analysis for that

I uploaded the SGF to my library:

I think this would be fairly straightforward to make, and in a sense, it kind of already exists.

I have not used KataGo, but I’ve heard that it is capable providing scoring and territory estimates, so it seems like the hardest part of making this tool is already done. The only question is the quality of the bot, which could, in principle, be improved through more training.

Thus, I think if someone wanted to do this type of practice, they basically already can using KataGo, which I believe is quite accessibly bundled with Lizzie. The only remaining task is picking random game positions, which could be done by simply downloading random games from OGS or GoKifu, and then picking random positions.

How are you envisioning such a tool would be presented? As a desktop application or maybe integrated into a website (like OGS)? Since such a tool would just be providing a layer of convenience on top of an already existing tool, I think the latter case makes more sense. For the former case, I could see it becoming a feature added to Lizzie, but I don’t know if it makes sense as its own standalone application.

2 Likes

I think there’s a potential issue with this approach. For human “counting” during the opening and early mid-game, a lot of the board (particularly the center where not much development may have happened) might simply be omitted since it is so uncertain. Hence, often human players essentially only just consider counts of developed territory. However, if a bot is giving a general score estimate at some intermediate position in the game, then it I believe it might also be doing much more than a human count by considering the points that might still be left to be gained in undeveloped parts of the board.

Hence, comparing a human count (of a partially developed board) to the bot count might be hopelessly in disagreement.

1 Like

Counting is important for pros because their games are usually fairly close. Same for high dans, but perhaps not as often. However, counting can also be a dangerous distraction. Haylee lost one of her Live Go games by distracting herself with needless counting. IMHO, counting has limited value for SDKs and below becasue the games are not usually that close. I have counted a few times in my IRL games and found it useful for deciding whether or not to play it safe. Far more often I know without counting that I am comfortably ahead or uncomfortably behind.

3 Likes

I agree with what you are saying. But i also think that a counting teaching tool could be useful. But there is certainly merit in what you are saying.

1 Like

Yes, exactly. Precise counting is meaningless (until the endgame starts). There are quite a few pros that don’t even do more than the hand method described in my last post.

Edit 1: Developing a counting tool is difficult because the method you use to count varies greatly depending on the state of the game, the amount of time you have, whether you care to count or not, whether you want to devote brain power or not, etc…

Although, there are pros like Lee Chang-Ho that are so good at counting they can make wins come from nowhere. That’s why I honestly think the best method is the method you find to be the most enjoyable and works the best for you. I guess that applies to everything you study (or refuse to study) in Go…

Edit 2: I would suggest having something that works backwards from a finished game (something I’ve been wanting to write down and flesh out but haven’t). Everyone can agree on the final score of a completed game. Then you remove the endgame moves and “count”. See if one player should resign or not. Then you remove a bit more and go to the end of the mid-game, and work backwards until it’s meaningless to count or estimate. Estimating (not counting) is also a skill, and is basically what the hand method is.

3 Likes

I kicked off a full strength AI review for you :slight_smile:

1 Like

I don’t think the style of katago’s estimation is very different from a human’s, she’s just more accurate.

AFAIK she has no problem scoring pro level games and as such should be fine for all human use (perhaps she might not be as reliable for bot vs bot games where both bots are stronger than her, but that is not the topic at hand)

1 Like

It is not uncommon for me to think the game is leaning heavily one way or the other, and for it to actually be much closer than I assumed. But certainly I am not advocating for counting in every game, simply that for those who wish to count better, it would be nice if there were a tool to help that.

3 Likes

What I mean is that KataGo will (by default, I suppose) count the whole board, while humans would naturally only count part of the board, if the game is in the early stages.

For example, see @Kaworu_Nagisa’s earlier post, where the center of the board is left uncounted, since from our human perspective it is very much up for grabs. However, an AI bot given the task to read as deeply as possible would attempt to count that as well.

It could still be a useful tool, but the humans using it would have to be instructed to count to whole board, at least guessing what the center is worth for each player.

1 Like

KataGo could be good enough to work out what “will be” vs what “can be” with regards to potential territory and influence.

If so, then maybe we can reverse engineer something from KataGo and update our human estimating/counting methods. Would be cool.

2 Likes

Actually KataGo bot draws the lines at more or less exactly the same points as Kaworu…

You can see that the darker squares mostly line up with the areas he said were likely territory and the lighter squares mostly line up with the same areas he said were influence.

3 Likes

Interesting, what is the count that it produces for that position? It seems like it assigns some of center with a light black. Does that mean that KataGo is giving some more value to black in the center?

1 Like

I think it said black up by roughly 10 at this point but I didn’t include any captures I just brute forced this exact position

Out of curiosity, what does it say for when white invades at S15?

1 Like