How pros count the score

[Krama]

Somewhere, and I can't recall where (probably on KGS lecture) there was someone showing how professionals count the score fast and easy. It was something like count two empty points as one and a dead stone as one or something like that... and then what ever you get just divide by 2.

I probably messed up with the description but it was few months ago and I really can't remember. Does anyone have any idea what I am talking about? And if yes could someone show a simple example, or perhaps even a better method for counting?

[xed_over]

Somewhere, and I can't recall where (probably on KGS lecture) there was someone showing how professionals count the score fast and easy. It was something like count two empty points as one and a dead stone as one or something like that... and then what ever you get just divide by 2.

I probably messed up with the description but it was few months ago and I really can't remember. Does anyone have any idea what I am talking about? And if yes could someone show a simple example, or perhaps even a better method for counting?

sounds like our very own DrStraw:
http://senseis.xmp.net/?SteveFawthrop%2FCounting

[Krama]

Yes, I think this is it!

Thank you kind Sir!

[moboy78]

Pros count things the same way we mortals do: 1, 2, 3 . . .

[Aidoneus]

I thank theres gold in them thar hills, an Imma a fixin to get out ma pick. Thank you DrStraw!

[RobertJasiek]

how professionals count the score fast and easy. It was something like count two empty points as one and a dead stone as one or something like that

Supposing they have already determined the intersections to be counted (which is the slow part of positional judgement), counting pairs of two empty points is neither fast nor particularly easy (although Cho Chikun describes it as his "high speed" method). Every really fast method of counting known intersections should include a counting of rectangles, of which one is added per dead opposing stone and one is subtracted per own stone inside a player's rectangle. For counting the REMAINING intersections, which do not fit well in rectangles, one can choose among a few additional methods, of which I have seen none described by professionals (except for the relatively slow "count in pairs of 2").

Besides methods for counting a 'count' (which is the techically correct word until the game end, when one counts the 'score'), it is even more important for speed of counting, to know and apply methods for UPDATING counts. These include the total count, either player's count, each region's count and recalling which have been the previously counted intersections. Again, I do not recall descriptions of such by professionals.

Of course, we can expect one professional or another to use at least part of such methods, but your request for how pros actually count the already identified counting intersections (and prisoners) is not easily answered, because there appear only a few descriptions for how they do it. So why do you ask only for professionals? Amateur advice is much better, AFAICS from the English or previously reported Asian literature.

(Your question does not appear to ask for how to identify the intersections to be counted, or how to do it quickly. Please clarify if you want to know also about that and why you'd ask only for professionals' methods.)

[oren]

Of course, we can expect one professional or another to use at least part of such methods, but your request for how pros actually count the already identified counting intersections (and prisoners) is not easily answered, because there appear only a few descriptions for how they do it. So why do you ask only for professionals? Amateur advice is much better, AFAICS from the English or previously reported Asian literature.

There are quite a few descriptions of how they do it. They just don't happen to be in English.

In a simple way, smaller areas are usually in pairs until you get to large areas and see rectangles used.

[RobertJasiek]

There are quite a few descriptions of how they do it. They just don't happen to be in English.

Nice, so what are those methods?

In a simple way, smaller areas are usually in pairs until you get to large areas and see rectangles used.

Even when rectangles are inapplicable, one should first consider counting with bigger multiples than 2. E.g., counting in multiples of 10, 5 or 4 can be faster than counting in multiples of 2. OTOH, when the shapes are very irregular, I find counting remaining single intersections faster than counting pairs of 2. Besides, there is more basic advice (of which I have seen none described by professionals, not even in diagrams): proceed region by region, within each region proceed methodically, consider counting first the empty then the occupied intersections (or vice versa), under area scoring and until the late endgame count the territory count (except for territory in asymmetric sekis, one sided dame and handicap stones).

[Mef]

I can't speak for the methods of pros, but one thing that may help you if you have a problem counting the board quickly is avoid the problem altogether. If you know that when you settled a corner it had ten points, it will still have ten points up until someone plays there. No need to recount it every time you assess the board. If you are making incremental changes to counts you have already made as play progresses, then it will only be the very rare case (such as considering a large trade) you must count anything quickly.

[DrStraw]

I can't speak for the methods of pros, but one thing that may help you if you have a problem counting the board quickly is avoid the problem altogether. If you know that when you settled a corner it had ten points, it will still have ten points up until someone plays there. No need to recount it every time you assess the board. If you are making incremental changes to counts you have already made as play progresses, then it will only be the very rare case (such as considering a large trade) you must count anything quickly.

This is part of the lessons I created and I am pretty sure it is the way pros do it.

[Abyssinica]

Why are we always concerned with counting "quickly" or "easily" when we can just take our time to count. I do the by two's method and don't feel it takes too long.

[John Fairbairn]

There is no fast and easy method. It takes time even for a pro to do a careful count, and so they do it rarely. Even when they do make a careful count, there is no special method. Personally I have never met a pro who uses the pairs method, though I've met some who've heard of it. Each pro has his own preferred method.

The most detailed exposition I have seen was in conversation with Takemiya. He showed me how he rearranged the stones within an area (mentally, of course, though in the demo he moved the stones) so as to form rectangles or other shapes the scores of which he often knew just by looking at them. In a simul against amateurs he would assess very large rectangles (a common feature of his style, of course) by shading the area with a hand at a standard distance and assessing whether it was above or below the hand outline. Extra points (e.g. if the area had 11 instead of 10 points) were dealt with in a different way according to the game. In a quick game he would be satisfied with 10+ or 15-. In a careful game a separate tally of the outliers could be kept.

One important feature was to offset areas against those of the opponents, and so drop them from the count, and this could be used also to deal with the stray extra points.

As to what is counted, a distinction is made between the solid territory and the boundary. Incidentally, this is a good reason to use the superior term boundary play rather than yose as it focuses your mind on the boundaries. The boundary regions, especially at the corners and sides, tend to fall into easily identifiable patterns and a pro will know the (deiri) value of many of these by heart - for some, well over 1,000 I gather - so he just superimposes a known count onto the solid count to get his estimate.

Captured stones can be dealt with in different ways. The idea of pair counting is one way. Another is to keep a separate running tally of prisoners. If I remember correctly, Takemiya would use captures as a way of making areas mentally regular but I also seem to remember this as feeling like witchcraft. As I understand it, one method never used is to count the stones in the bowl lids. For a pro, who has to give many simul games in his life in messy environments, this would be fraught with danger, and in slow title games would be just plain infra dig. Instead, they look at the position and remember from the shapes where captures were made. An anecdote somewhat related to this is from a Japanese tv game in which the player adjudged at the end to have lost by 2.5 was amazed whereas the opponent and pro observer seemed to think it was ok. It was only when the player leaned back in disappointment that he noticed three prisoners in his lid that he had forgotten to put on the board.

Of course, the mental updating method is pretty standard. It has to be, as time runs short at the end of games.

I've been told that the area method (using Japanese-style komi) is normal during a game in China, as counting the stones as well is just too inconvenient. And for them, counting the prisoners in the lid is not even an option as they tend to be tossed back into the bowls.

Whichever method is used, none is clearly entirely reliable. We can tell this even from published commentaries. There are many, many cases of pros getting the count wrong and making wrong strategic decisions. Clearly, they are often using the updating method and building on a systematic earlier mistake, because if they re-counted in full they would probably spot the systematic mistake. A common such mistake, apparently, is to assume a capture where there has been none, e.g. a ponnuki shape where there was no actual nuki. In Korea, there have even been cases of applying the wrong komi (mentally).

There are also cases where players have disagreed over the final count. The most famous was in a game between Kitami and Hashimoto. Admittedly it was a fast game. When the game was first counted it came out as a win for Black by 0.5 point, but Kitani (White) challenged this. The game was replayed several times by the commentator, announcer and scorekeeper (all pros), and each time it came out as a White victory. Hashimoto eventually accepted the loss.

There was a Tengen title match game (i.e. slow play) where the players and pro referee all agreed on a final score of W+0.5 but it was really W+1.5 (as discovered when the game was input into a computer, though one player discovered it himself later at home).

If pros find it hard, are amateurs who offer quick-and-easy and/or reliable solutions snake-oil salesmen? Well, let us not overlook the ultrafamous Ear-reddening Game, for years assumed to be B+3 when it was really B+2 - as discovered by an amateur.

[RobertJasiek]

Why are we always concerned with counting "quickly" or "easily" when we can just take our time to count.

If it takes ca. 2 to 3 minutes to determine afresh the intersections to be counted in one position, it is easy to see that one cannot do so for each position of the game and each position created by each imagined variation. Methods for updating counts etc. are needed so that one can do good positional judgement for every important position.

Since it can already take 2 to 3 minutes to determine afresh the intersections, it helps to count those intersections as fast as possible. The time gained can then be used for other things, such as more careful life and death reading.

a good reason to use the superior term boundary play rather than yose

You have suggested this for years, and I like the term boundary play, but for a different reason: it is a more general term than yose, because the latter is associated too much with the late phases of the game. Boundary play can occur during every stage of the game. As a different special kind of boundary play, I invented (or maybe used again) the term mutual settling fight for those fights in which boundary plays create, or settle the basic outline of, a boundary, as it can happen during the opening or middle game. All boundary plays / fights have a common feature: either player also aims at optimising the value of his territory.

I've been told that the area method (using Japanese-style komi) is normal during a game in China, as counting the stones as well is just too inconvenient.

(Convenience depends on the shapes.) As others and I have explained elsewhere, territory counting can be used under area scoring until the late endgame, because the territory count (almost) determines the area count (except for adjustments in asymmetric sekis, one-sided dames, uncompensated handicap stones).

Whichever method is used, none is clearly entirely reliable.

Of course, but good methods (such as mine) should approach 0.5 points accuracy for 'quiet' positions, provided that territory count and the turn are the only relevant factors of a position's positional judgement. If influence, severe aji, strategic choices etc. have a great impact, then a single number may be insufficient description.

Clearly, they are often using the updating method and building on a systematic earlier mistake, because if they re-counted in full they would probably spot the systematic mistake.

IMO, you overestimate such mistakes by pros; I would expect most of them to do more PJ than you suggest. However, they can fail if territory is not the only factor.

If pros find it hard, are amateurs who offer quick-and-easy and/or reliable solutions snake-oil salesmen?

Quick and easy is only the counting of KNOWN intersections. (Despite your fun trivia counter-examples.) The time needed for their determination enables greater accuracy, and I recommend to use at least 90 to 180 seconds for a fresh positional judgement done by experienced players.

For reliable solutions, this very much depends on how quiet, stable and settled a position and its components are and whether other factors besides territory and turn play a decisive role.

[John Fairbairn]

I've been told that the area method (using Japanese-style komi) is normal during a game in China, as counting the stones as well is just too inconvenient.

I think, from what Robert said, I've mixed up 'area' and 'territory', but the distinctions of the rules mavens bore me anyway. In normal English, what I mean is that during a game Chinese players count the same was as Japanese and Korean players.

[RobertJasiek]

Yes, and this is essentially correct if applied only until the late endgame (when the Chinese should consider area counts more) and if modifications due to asymmetric sekis, one-sided dame and handicap stones are not necessary (if Chinese (or foreign pros in Chinese tournaments) do not consider these, they are making mistakes).