Lately, I've noticed a tendency in my go that might be familiar to you. It began dawning on me in conversations with Skydyr and reviews from some nice folks (Frozensoul, Starstorm etc.) on KGS. With the moves you play you try to squeeze the most out of a position as possible. This is because of our shape- and tesuji-knowledge. However this always leads to my games being complicated (big exchanges, leaving thin positions because you want to keep sente). I don't know for how many of you this 'habit' is familiar.

The actual topic here a) counting and b) adapting your strategy based on the evaluation of the count, which could be a solution to this challenge. I found this page on Sensei's: http://senseis.xmp.net/?SteveFawthrop%2FCounting. On the page are excellent lessons in .sgf-format which explain how to count in your games.

Estimating the score in various stages of the game helps me in a major way deciding whether to invade or reduce; exchange influence for territory; backing off or fighting. Visualizing different variations and how these influence the score also can help us evaluate the results of candidate moves.

For myself, I think counting will help me play more calm and less envious. I find that I keep the score difference unconsciously in my head and find the sequence that gives the amount of risk/reward needed.

P.s. The lectures are certainly worth checking out in the SDK-region. The precise methods Steve Fawthrop indicates are not my preference. For example, I do count in pairs but I count in 'two's' (not 1,2 but 2,4 etc.), instead of what the lessons recommend.

Territorial positional judgement should be 95% determination of which intersections are a player's territory and 5% counting them. One can count anything, but if the count is 5 points off because of having determined wrong intersections to be counted, one can easily make wrong strategic decisions.

Why is counting in 2s so popular?! I recommend a combination of counting rectangles, adding dead stones and subtracting friendly stones from rectangles, counting in multiples of 10, 5, 4 or 2 whenever a remaining part cannot be counted well in rectangles, possibly counting empty / occupies intersections separately etc. Combination of such can accelerate counting by a factor 10 in comparison to counting 2s. Besides, it is very useful to use incremental updates of one's counts and recount only the changes.

Steve Fawthrop's counting lessons are not excellent, but contain some useful and some dubious advice. E.g., his first advice to always count in pairs is dubious, see above. Much more efficient techniques are possible!

I dislike his concepts "probably points" (counted as full points) and "possible points" (counted as half points). While counting half points can be a useful means, it must be applied to meaningfully chosen intersections such as gote options or moyo options. Instead of teaching kyus rules of thumb (about probably or possible points), they are better taught the same as dan players.

His advice to count only in multiples of 5 in the fuseki and calling everything more precise "impractical" is bad, because counting in the fuseki can be similarly accurate as in the middle game. In particular, counting in multiples of 5 implies an error of up to 2.5 per counted local region. Very easily, such errors can increase for several counted regions; they need not accidentally cancel each other.

His (or John Stephenson's) "good way to define thickness", that is to look at the stones and ask whether they will be uncomfortable if the opponent comes near, is very imprecise, because "uncomfortable" is nothing measurable. Instead, I suggest to use my definition of thickness and consider the measurable aspects of connection, life and territory potential.

He claims that influence and thickness were hard to evaluate. This is not the case, unless one wants much greater precision than needed for positional judgement. In particular, one often assess (greater) influence by the Influence Stone Difference. Maybe it was hard to evaluate when he wrote his SGFs, but now reasonably easy methods are available.

Counting points within a distance of 2 or 3 (full) or farther (half) of thickness is a bad method. I understand why he proposes it: it is one of the first things one imagines when trying to find a general method. I started with such a kind of method in the 90s for myself. However, pretty soon I disregarded it again; it is too much rule of thumb and imprecise. Instead, I recommend consistent usage of "current territory", regardless of whether it is in safe territory regions or in front of a thick wall. Thickness should be evaluated as thickness, not reduced to its local territory generating impact.

His suggestions to re-evaluate counts and base strategy on counting and its (im)balance are good.

Robert, I cannot dispute anything you say - for a high kyu or stronger. But please remember that these lessons explicitly say they are aimed at low and mid kyu players. What I tried to give in those lessons was some very basic ways to estimate the score for someone who has no idea what counting it about. They are intended as a starting point for people who can develop their own methods as they improve.

Given that, I stand by what I wrote. And remember, those lessons are ten years old and have retain popularity all this time. They obviously fill a need.

_________________
Still officially AGA 5d but I play so irregularly these days that I am probably only 3d or 4d over the board (but hopefully still 5d in terms of knowledge, theory and the ability to contribute).

Agreed wholeheartedly - I remember finding the counting lessons quite helpful (I still misjudge score by 20+ points, but the fault is purely my own - the lessons certainly opened my eyes to another way of thinking).

Counting more has helped my game become more like my idol's, and has definitely improved it. 'Less envious' is definitely the feeling here. I now tend to gently approach or reduce many positions that usually scare players of my (effective) rank by their size and influence. Surprisingly, I get my inspiration from Gu Li and Lee Sedol. I find them using this sort of gentle approach leading up to the middle game.

Counting and simple reading can definitely help to decide how much risk you need to bear. I'd imagine Honinbo Shuei and Takagawa Kaku not only reading out the probable continuations from a position, but calculating the most likely local outcome at the end of the game, and decide their course based on that, in part.

Strong players count. That's all there is to it. If you want to become strong, the best thing to do is start early and keep up the habit.

I've recently begun counting, and it definitely makes a big difference. Very often I am surprised to find I am in the lead, where I felt the reverse was the case, or vice versa. The problem I face is that I seem to really battle to maintain the score. I do a rough count, which takes me a fair while, and then, 5 moves later, I am all at sea again, and time often doesn't permit a recount. In games where there are large captures, I am completely lost when it comes to a brief count. Never do I get to the end of the game and think, "aha! Got him by 6 points." It's always pretty exciting when the score is added up.

Quite why what should be a relatively simple procedure seems so difficult in the heat of battle I am not sure, but I will keep trying, and I assume that it will get easier with time.

Tygem has a Score Estimator that is available to players during the game, but I try not to use it before I do my own count. I have been known to check it pretty often toward the end of the game though.

Counting by twos is so popular because it is so much simpler than the method you suggest. It requires virtually no calculation and you don't need to decide for each shape which rectangle would be best. As for using incremental updates, why should rectangles be any better?

_________________
Patience, grasshopper.

I have found that counting in pairs is the only way I can be confident that I've counted dead stones and captures correctly.

I count by 1s. There are days I think I'm the only person on the planet who does that. I've tried counting by 2s or larger, but maybe I didn't play enough tetris in my youth.

I also count in 1's. I guess I prefer it because that's (a) the normal way to count and I am less likely to lose track of where I am and (b) I'm not confident enough that I can find pairs of territory in complex areas. Counting in 2's is compelling because its theoretically faster and its so much easier with prisoners, but I never got comfortable with it.

Practice and you will. Use the subtraction method from my lessons and you will never look back.

_________________
Still officially AGA 5d but I play so irregularly these days that I am probably only 3d or 4d over the board (but hopefully still 5d in terms of knowledge, theory and the ability to contribute).

DrStraw, unlike you I think that weaker players should learn (thus far only the basics of) the same of what stronger players should learn. This has the advantages of already reasonable approximative accuracy and no need to unlearn dubious advice later.

daal, concerning popularity, from my observations, counting in rectangles where possible is more popular than strictly always counting in 2s. Concerning simplicity, IMX counting a big rectangle is much simpler than counting its intersections in 2s, although the method becomes less simple when allowing a few basic operations ('counting a rectangle', 'counting another pair of 2 or half of it' etc.) instead of only one ('counting another pair of 2 or half of it'). The calculation of counting a rectangle is elementary school level. One can restrict counting rectangles to those where their choice is obvious.

Incremental updates in themselves do not require counting rectangles, but the latter and further basic operations greatly accelerate counting compared to always counting in 2s. The greater speed of full counting and the speed of incremental counting itself enable frequently performed territorial judgement, such as doing it on every turn.