Counting practice

First question: does something like this exist?

Second question: if not, how hard would it be to make?

So I was thinking, and it seems there are tools to learn almost all skills except counting.
My idea was something like katabot could generate random board positions… You’d give it an integer value for the difference of scores (maybe positive for black winning and negative for white winning, only the board state, ignoring komi and captures) and it would return how close you were to its own estimate, and allow you to generate a new board if you desire.


  1. very unlikely

  2. I don’t think it would be hard, if the score estimator was good and vetted by strong players.

But it depends entirely on whether you want to estimate the actor in the midgame or endgame, or after both players have passed.

All three are possible to do, but require veery different skill sets.

I can help if you want.

1 Like

The other day I started spectating a game and trying to count as much as I could as they played. I ended up with a super abbreviated syntax for recording the count.

<upper right> [.<right>] <lower right> [.<top>] <upper left> [.<left>] <lower left> [.<bottom>]

Bracketed are optional, to omit non-bracketed entries (corners), put a comma immediately before the next entry. Use a semicolon to separate Black from White.

In my effort to learn counting I have divided the board in 9 parts, easy to guess. I found quite easy to count each one but very hard to keep counting from one to another.
Also I try to recognize shapes to avoid counting all points one by one, like 5x2 or the bulky five or 3x3 and so on… but I never found someone teaching that

1 Like

I like the idea :smiley: Once I have time, I would consider creating something like that for the tutorial.

Not sure about creating something that could auto-generate those positions, but creating some basic database would not be too hard at all. Just a bit time consuming…

I saw at least two videos on counting?

What I do:

  • count by pairs
  • square-ish areas can be counted (which I love doing) via just a diagonal and then squared. For rougher estimates that is super fast to do.
1 Like

I must look for them.
Any link?

Got one by Nick, can’t recall the others, sorry

1 Like

There are some good clips on counting available on Sesame Street:



You could have the bot play a game against itself to generate a reasonable board position, or maybe it could just pick a random position from a library of games.

I’m guessing that you mean more broadly the positional evaluation at any point during the game, and not just counting the score after both players have passed and agreed on stone status. Is that correct?

1 Like

Yes, absolutely. I would see practicing scoring the final position of the game after both players have passed as borderline pointless as in such a situation you have no time pressure and are able to move stones around.

My idea was to have various positions in a mid-game or late-ish end-game state to practice score difference estimation compared to a reliable estimation (for instance, kata bot)

Dan players keep telling us it’s important to count during the game (well, except dwyrin :smirk:) , but short of “just doing it” there aren’t really any ways to practice this.

1 Like

Ironically, my biggest problem is remembering the first player’s score estimate by the time I’ve finished counting their opponent.

A tip I have heard only once, but surprisingly nowhere else, is to count up one side and down the other.
Eg. if I count black to have an estimated 64 points, rather than counting white up from 0, I’d start with white at 64 and count down. I’m still no less likely to remember either total but it theoretically (still learning / practicing) is an easier path to get the score difference which is what really matters anyway :slight_smile:


The only real thing that I see that is missing from suggestions on here (Unless I have misread or skipped something) is that the ideas presented offer no time pressure training.

There are two main ways to count a game… and both are valuable in different stages of the game. There’s the educated guess and the actual count.

But there is something that links them… time.

In different parts of the game, it is either easier or more difficult to count the Game. After the fist fifty moves in a game, the count is an educated guess over which slices of the pie are going to get points. But it’s just a guess. But later in the game, there is a more certain count as there are fewer variables to work out, and the count becomes more accurate naturally. But time pressure is something that can really skew how well a person is able to count.

So my suggestion would be this… Three or more stages. Each with time controls.

Beginner: over a min to count (The board is a completed game)
Intermediate: thirty seconds to count (most endgame is incomplete)
Advanced: between 15 seconds and thirty seconds depending on the board. (The board can be any of the above, and indeed should include games still in the midgame. Because knowing if a player is ahead r behind can change the type of moves that a player has to play to either secure the lead or to gain the lead from the opponent.)

Of course, there could be alterations to my idea here, as I have never developed anything like this in my life… and this is just an idea. But I think adding time pressure could very well be a great way of actually training people to count in a more realistic way.

In terms of the randomly generated board… I don’t think it should be all that random… There is already a database of all the games ever played on OGS. Some games are completed games, some are not. My point is that we already have the resources to support such a thing being created. And the boards need not be all that well played out. As in… a low kyu game could be as valuable as a high dan game, for counting, because of the different situational problems around the board.

There is also the added issue of prisoners. This is something that should be factored into the counting as well I think. So some boards have captured stones to be added to the actual count of the board. This adds in a complication that all players should already be fairly familiar with, and it’s realistic to have this in there as well.

1 Like

I think this is possibly a bad idea for the same reason that I’ve never seen timed tsumego puzzles (they may exist, but I’ve not come across any).
I believe that during the process of training, accuracy and effort are worth more than speed. The idea is to train the muscle. Even pros and high dans will often spend many minutes counting the board, I don’t think it’s reasonable to expect novices to do it in mere seconds.

As you get stronger at counting, you will naturally get faster at it. Adding an extra expectation of speed from the beginning just gives people an extra reason to quit, IMO.

For a more advanced mode, I think a stopwatch might be a good idea… so that people can try and beat their best times, and I think that lends itself more competitively than a simple pass/fail countdown.


I like the stopwatch idea. But I also like my idea. I don’t think I have the times absolutely right in my idea at all, but they were examples.

I do however have an added suggestion whereby you could choose to be timed, work within in a time, or have no time at all. I don’t think it needs to be any single option, I think having multiple options is a great way forward.

Also i have see timed go problems. @xhu98 and his teacher have done them live on stream, but the site they were using was in Chinese I think??? but other than that that is the only time I have personally seen timed go problems.

I do still think that having the option to be pressured by time, maybe an attractive option to some players. But having said that, I do also completely understand what you say, and I do partially agree with it also.


“Clock is the worst influence the west has had on Go” - Tokumoto

:100: :clap::100: :clap::100: :clap::100: :clap::100: :clap::100: :clap::100: :clap::100: :clap::100: :clap::100: :clap:


Also I never mentioned, I use the “hand” method of counting in the opening and mid-game. For example…

Areas A, B, C cancel each other, and whites potential is less than blacks, so white should invade on the right and not play to build, as white would lose the building race.

1 Like


1 Like

I may not be a pro go player, or even a particularly strong go player. But i have to dissagree with the whole time being introduced to go, as a bad thing…

With the introduction of timed games, we have seen a surge in the popularity, and access to the game, which has allowed such things like Alphago to come in to existence, and as such we have learned more about the game of go in the last three years than we have in the last fifty years easily… possible more like a hundred years.

Because we can have quick timed games, means that the game is accessable to those who dont have several days to complete an intense game of go like they did back in the day.

Has the quality of go decreased from having time controls? Id say, overall … no it hasnt.

There is of course the danger of Pro games becoming too short… and in my opinion the Transatlantic Go Championship games were too short. I think three hours main time with one min overtime (Buyomi) is the best time control for a managable game in terms of length and quality of play.

Also, i quite like youe Hand Method. Its sort of simmilar to what i do now… where i calculate the average size of areas to suggest who is ahead and what they have in terms of potential vs points. I shall have to factor in the whole “Canceling out” part though to see if that makes my estimations any better, whether it be faster or more acurate.


It’s faster and lighter on the brain power, because you never have to think about those areas again.

I think the popularity of the game has more to do with accessibility through platforms like OGS and KGS, and news like AlphaGo rather than simply having a timer. Timers have been around for a while in Go, long before the internet and AG. It didn’t do anything for popularity.

Timers are good for having a short game, but you’d be wrong to think that every game was days long before timers. All of the games in my old Go club were palyed without timers, and we still played ~30min to 1h games, even without a timer.


I don’t know the history of timers in Go but it would have to be inevitable for online play. Rather blame the internet than the West.

I have also enjoyed (tournament) club games without timers. For what it’s worth, I find fischer clocks much better for allowing a game to have a natural feel to the pace. ps. Byo-yomi doesn’t sound West to me.