Any suggestions? I frankly have no idea how to estimate my score, even at 9k!

Have you tried guessing?

Ha ha. Unfortunately my guesses are even worse than the bad AI estimate.

ThatĀ“s why I am asking. What do the Dans do? They must have a method of estimating their score. Or are they doing a precise count every couple moves?

You donāt need to count every couple moves, but only when you need to make a strategic decision like invade vs reduce.

There is a book on this, by kiseido/ishi press.

But to summerize a bit:

Count only sure territory. (To be very minimum size)

Count 2 by 2 is easier (a prisoner is 2 points btw)

Then to have a more fine evaluation you need to study the endgame, there are books on this too.

The answer is in your question. You need to estimate and you need to be fast.

Just to find out which territory will be black and which will be white is a challenge. If there is an isolated border section, you close it off in your mind in a ābalancedā way. For example, if it is gote endgame for both sides, you assume that both just ādrop downā to close it off with an equal amount of stones, even though the real endgame will go one way or the other.

If you have a framework, you use your judgement on where to close it off. You can use the āsector lineā, or assume the most severe sente reduction where the defender always responds passively.

Having made up their mind on the borders, most strong dan players count in pairs. Territory usually comes in shapes where points are right next to each other, so you only count half and then double it. I personally use block shapes of 15, 10, ~4-6 and singles, which is a bit slower.

If there is just one very uncertain section, you can skip counting that, call it X and use this approach: āX needs to be 20 points to be equalā. If both have frameworks, you can also use a formulation like āX needs to be 10 points bigger than Y to equalizeā.

Also use your experience to tweak the result. For example, if there is a weak group under attack, count it as 2 points, but assume that the player under attack cannot invade or reduce like normal.

Most importantly: practice counting a lot! Donāt put it off until you are 1-dan. It will improve your Go.

You can compare territories by size visually to get a rough idea. It would go like: āthis blackās corner is similar to this whiteās corner, white has this big side but black has these two territories which combined approximately equal to whiteās big side territory.ā Make your guesses on whoās ahead by how much, write them down. After the game compare them to ai graph. If your guess differs drastically from aiās try to find where your judgment went wrong. Then repeat. With practise youāll become better at it.

Counting is a very strong tool, but I think you donāt really need to worry about it until you reach 4-5 kyu. Before that judging by eye and comparing sizes is sufficient enough.

Also not using score estimator is best to grow your estimation or counting skills.

Also another thread like this one

I donāt actually count as much as the received wisdom suggests you should. My rationale is:

- In middle-game when thereās still a lot undecided, letās say my counting accuracy is +/- 10 points
- thereās only value in counting if the result will make me change how I play, e.g. am I ahead and I can play a gentle reduction, or am I behind and I need a deep invasion?
- if I count and the result is me +5 then with +/- 10 error bars, I cannot reliably act on that information to decide how to play as maybe Iām actually -5.
- if I count and the result is me +20 then I can trust that, but if I am ahead that much I would know that just form the game flow feeling and not need to count, so counting is pointless.

Of course my ability to judge who is ahead from game flow feeling is itself a skill which will be better than yours. This is positional judgement, knowing what fair results look like and spotting inefficiencies in the resulting shapes and basically like inertial-navigation: the game starts even, I make a small mistake, opponent makes a small mistake => still even. I make 2 big mistakes, he makes 3 big mistakes => Iām ahead.

I prefer counting only the secured points for both sides , it will help next when refining by adding endgame sequences. And it avoids some overestimations.

If an endgame follow-up is (truly) sente for one player (say black), but gote for the other (white), then you should count the points with black getting the sente endgame, and counting only the currently secured points will lead to an underestimate for black. Itās quite possible for these sente endgames to not roughly cancel out around the board, but one player to have the privilege to get lots of such endgames (usually the one with fewer groups who played a thicker game), so counting only secured points can easily lead to being tens of points wrong in a count.

I didnāt say itās enough, itās just the right base to start evaluating the endgame then. Yose is complex with assessments of the value of the threats vs sente/gote, a crude picture is to compare what each owns exactly and assume a balanced endgame between both. If you start to integrate apparently sente sequences then you have to go all the way for each spot to be fair.

I never count, I donāt think counting is necessary.

Almost 99% of my games, I either win 50 points or lose 50 points.

Whatās the point of counting in these games?

I rarely come across a game where the difference is within 10 points, so I donāt think itās necessary to spend time studying it.

I think knowing how to kill your opponent or survive yourself is more important than counting.

There should be something wrong in the way you manage your games.

If you take games played by experts, itās not working like that. To kill or not be killed is surely kind of essential but you wonāt see this happening so often, itās used more as a threat as really happening. And then with these you get a numerical advantage for a win.

Experts? you mean dans?

in sdk, groups still often die for no reasonā¦

O Iām not talking about immunity but how their games are following a general pattern.

I had a recent tournament game with super long Fischer controls where at some point while fighting the last ko, I wanted to know (a) do I need to fight this ko, (b) do I need to try and make the ko bigger, or is winning the ko enough.

I took some time, counted the game slightly incorrectly it seems, in that I thought if I win the ko I win by 0.5. I miscounted a move I thought white would need to play defensively later. In any case I knew the game was close and I needed to win the ko.

My opponent also counted the game, but rather than have a numerical value, they just thought āitās close, I need to win the ko to be safeā. I think I also had this kind of thought, but I wanted to count more precisely because I had the time, and also if I was just losing by like 5 points maybe we could just finish up early

In any case it turns out white was winning by 0.5 even with black winning the ko, but White played a non working ko threat in the black area. Black connects the ko, and gains one extra prisoner. I thought Iād win by 1.5, because of the miscount, but actually won by 0.5.

In any case with a more precise count white couldāve even passed and won by 0.5.

I think that might be most useful bit of counting Iāve done

In other games I did try to make some decisions based on counts, but depending on the time settings, and limited by my own abilities, it sometimes just didnāt work out.

Some other recent games however in a separate tournament I was able to estimate a rough lead and then just make decisions that seemed safe enough to reduce areas, play for endgame, and turn certain areas into dame and be happy with that.

Counting can be useful, but accurate counting and accurate play also need to accompany it in some cases.

Can you remember the title of the book?

Thank yĀ“all for your insight.

Follow-up question:

Being lazy, I use this method to estimate who is ahead:

**This is not advice. I am not nearly good enough to give advice. ItĀ“s just a description what I do.**

I count a corner as 20 points, every sideline as 30 and the center as 40 points.

So if I control 1/3 of a sideline I estimate this as 10, half a corner 10 again and one quarter of the center another 10.

If I think I am ahead by 10 or more, I play defense. If I am behind, I try some dubious experiments to confuse my opponent. If about equal, I try to make the best move.

So, am I a genius or an idiot?

I think that thatās important not to count more territory than you actually have. If method you describe is as vague as I think, then you should not play ādefensivelyā if you assess 10 points lead and instead you should play normally, because margin of error might be around 10 āpointsā. If your lead is smaller than you assess, then playing defensively(as if youāre ahead by a lot) might easily lose you a game.