Area or territory – The future of go rules

Published 14 Aug 2020 by antti (last edited 14 Aug 2020)
tags: tldr, rules

(The historical information in this article is based on The History of Go Rules by Chen Zuyuan.)

As we saw in the first half of this post, go was first played with stone scoring, then with territory scoring, and then also with area scoring. Although territory scoring is nowadays attributed to the Japanese rules, they, similarly to the two other scoring methods, it originated in China. Therefore, in a sense the dichotomy of ‘Japanese rules v. Chinese rules’ as ‘territory scoring v. area scoring’ is in fact false. From a historical point of view, Japanese go players took in the Chinese game of go and its rules and built upon them, resulting in the Japanese rules of go. Then, in response, the Chinese of the 20th century built their own official Chinese rules of go to compete with the Japanese – even though both rulesets describe the same game, and the Japanese rules even have a Chinese origin.

While international politics may explain at least part of the rules competition between China and Japan, there is another angle to this topic. This is the fact that there are merits and demerits to both rulesets (and to territory scoring and area scoring), and neither is absolutely speaking superior to the other.

Territory scoring, and with it, the Japanese rules, are extremely efficient when it comes to scoring a game after it has ended. The territory-adjusting process is quick and smooth, and the whole-board position is also retained in the process, making the result possible to verify and also helping the players discuss parts of the game with the remaining visual cues.

A 19×19 game has ended. Black has captured 20 white stones, and White has captured 9 black stones.
In territory scoring, captured and dead stones are then moved inside their player’s territories, where they function as negative points. The territories are then rearranged into easily countable shapes, and finally the empty intersections are counted. In this game, Black has 36 points to White’s 33 points, so Black has three points more on the board; but when White’s 6½-point komi is added, White finally wins by 3½ points.

Area scoring, and with it, the Chinese rules, are more exact and straightforward when it comes to more complicated endgame positions. One example of this is the bent four in the corner combined with unremovable ko threats, which the Japanese rules unfairly deem dead.

This position was shown to me by Tuomo Salo, a Finnish 3-dan, a long time ago.
Under Japanese rules, the upper-right corner black group is deemed dead and the upper-side situation a seki. However, ‘a’ and ‘b’ are in fact unremovable kō threats for Black, so if White had to prove that the upper-right corner black group is dead, Black could be able to capture the four white stones on the upper side in exchange.
The Chinese rules are more fair in that they require White to start the kō in the upper-right corner. If the game ended as it is, all the black stones would be considered alive.

In exchange, the area scoring process is not as efficient or elegant, requiring one player’s stones to be removed from the board and the other player’s stones to be counted in a seemingly chaotic procedure. If only there were some way to automate this process … Oh, wait!

If the earlier game was scored with area scoring, the end result would look something like this mess. The black stones stand for ten points of white territory each, so White’s total score is 179 points before the komi is added.
But that fight on the left side sure was exciting, wasn’t it?

From a modern go player’s view, it seems like a false dilemma that go players should be limited to either territory scoring or area scoring all the time, when each method has their optimal use case. Namely, area scoring benefits greatly from the fact that a computer can perform the complicated scoring process automatically, without the board position having to be messed up or removed. Territory scoring, on the other hand, is as easy as ever to perform on a real board.

It seems to me, that rather than ‘Japanese rules’ and ‘Chinese rules’ (and the other numerous alternatives), whose names mainly carry the weight of history, we should have more descriptive ‘online rules’ and ‘irl rules’ instead. The online rules would make use of the computer's counting help and implement area scoring, thereby leaving the system’s greatest strength while eliminating its greatest weakness. As a valuable bonus, this makes it easy for beginners to start playing the game online, since area scoring can be explained off as stone scoring, which is arguably the easiest way to learn the game. The irl rules, instead, would implement territory scoring. Depending on the situation, if it was two beginners playing each other, they might choose to use stone scoring anyway; while, if they have a person explaining the rules (as is often the case in irl settings), territory scoring is then the easier of the two common alternatives for manual counting.

Beyond the choice of territory or area scoring, there are several more important choices to make for the ‘online rules’ and ‘irl rules’; for instance, whether superkō (and what type) is implemented, and whether suicide is allowed or not. As I am not an expert on rule technicalities, these details are probably better left for more qualified people.


Comments (6)


lightvector wrote 3 months, 3 weeks ago:

I think there are some important further considerations that this article has not taken into account that make the choice of area scoring not obviously ideal to adopt as the standard for a wide swath of games. I also feel like a few of the points in this article deserve a little bit more discussion.

One is that area scoring gives in practice a reduced granularity of scoring, incrementing by 2 points in the vast majority of games instead of 1 point. This has the consequence that it is not possible to set the komi in area scoring to be fair for high levels of human play. 7.5 komi appears to favor white. My understanding is that this not just for bots, but it is also borne out if one investigates high-level human game statistics (although one must be careful to restrict to sources where colors are assigned randomly as opposed to biasing towards giving White to the nominally higher ranked player) - and also I understand that many more pros than not feel that paying 7.5 komi is tough. But on the other hand, 5.5 komi is much too low and favors black. It turns out that 6.5 komi territory scoring is very close to fair, however!

Are you aware of Button Go? https://senseis.xmp.net/?ButtonGo .Essentially, you can use 7 komi with the rule that the first player to pass gets the half point. Ruleswise it has every advantage that area scoring does in terms of dispute resolution and teaching and ease of implementation, but leads to a finer and more fair game. It is not exactly equivalent due to the potential for ko-monstering the last endgame ko, but is very close at reconciling territory vs area and achieves the above goal quite well.

Additionally, I think you overstate how hard it is to count area in-person. Even if the way you describe might be one way that some games are counted in practice, it also misleadingly gives the impression that counting is hard and confusing than it needs to be because there are also less-confusing ways. I've counted many area games in-person with 19x19 boards, by the method of making the areas more rectangular for one or both players by rearranging stones to exchange points (quite analogous to how territory scoring counting would make the territories more rectangular by rearranging stones) and then simply counting widths and heights and multiplying, and also subtracting from 361 for the other player as a cross-check. The numbers are larger, so this is definitely harder than the counting for territory scoring, but it certainly doesn't "require" that one player's stones are removed from the board as this article claims.

Moreover, by means of AGA-style pass stones, one can even perform the area scoring by doing territory counting! So then all one needs to do is note how many passes there were and who passed last, and then the area scoring is trivially derivable from a territory count. So the claim that removing all of one player's stones is "required" gives the misleading impression that the counting process is very hard and confusing when actually there are multiple alternatives that variously are conceptually simple and/or barely any harder to perform than for territory scoring.

Lastly, if in-person games are going to be played with territory scoring, we are going to need SGF viewers and bots and analysis and scoring tools that handle territory scoring anyways, even in the online world, because we will want to reply and study and analyze those games! So keeping territory scoring available in the online world is still going to be necessary to some degree even in the best case, no?


lightvector wrote 3 months, 3 weeks ago:

Addendum:
I suppose the last post argued both "against" area scoring and "for" area scoring in different ways and contexts - mainly I think it's important to be accurate and consider things thoroughly before making any sort of radical proposal, as it seemed to me that there was both an unmentioned disadvantage of area scoring that applied *regardless* of online versus real life that was worth mentioning, yet also an overstated disadvantage of area scoring in the context of counting difficulty that also makes it not as bad as the impression that a reader might come away with.


antti wrote 3 months, 3 weeks ago:

Thanks for your comment, lightvector! I am sorry if the posts were brief when it came to details and backup arguments – I was trying to find a balance between ease of reading and technical content.

As for the ‘granularity of scoring’, I think there is a counterargument in that territory scoring finishes the game early when ‘originally’ (as per stone scoring) there should still be a last, often non-trivial endgame battle for who gets the last neutral point. I do give that the 6½-point komi under territory scoring seems to make for a fair game according to current AI research. Still, while the size of the komi is of course a topic to be solved, I do not consider it a deal-breaker in either direction.

Button Go seems interesting. I didn’t have enough time to think all its implications through inside my head, but it does seem to make the weight of the komi a little bit smaller under area scoring. Personally I would give it a small style penalty for the extra rule, however.

You are right that there are ways to perform area scoring effortlessly, probably the implementation of the pass stone rule being the easiest example. I now realise that in the article, where I had the diagram for area scoring, I should have written ‘Chinese-style area counting’, which my argument was mainly against.

I was certainly not advocating that the notion of territory scoring should be removed from online go altogether. What I had in my mind was closer to that, rather than have ‘Japanese rules’, ‘Chinese rules’, ‘AGA rules’, ‘New Zealand Rules’, etc. as options on online servers, go players (especially beginners) would benefit if the options were ‘online rules’ and ‘IRL rules’ instead – or possibly even just ‘area scoring (default)’ and ‘territory scoring’, with a few more hidden advanced toggles for superkō and the like.

The main motivator for my these two posts was not a strong opinion on what is the ‘most optimal ruleset’, but rather my having gotten tired of the false dilemma of ‘Japanese rules v. Chinese rules’. This may not seem like such a big deal in the West, but is ever present in Japanese go discussion (especially now that the AI came around). Because of related politics, even the IGF has been unable to adopt a universal ruleset, although it would be in an optimal position to do so, for the (arguable) benefit of go players everywhere. Therefore, rather than doing something more drastic, I meant to start by stirring up conversation among active go players. As you say, accuracy is important in this topic, and I’m not an expert on it either; therefore, I need the help of more knowledgeable people.


gerryg wrote 3 months, 3 weeks ago:

Some of us volunteered to participate in the EGF Chinese Rules tutorial. A problem identified with Japanese rules is the final ko and no need to fill dame. With a bit of effort I could dig up the actual pro example game which was won or lost on the requirement to fill the ko. AGA rules eliminate that problem. While I am sure it simply a matter of getting used to it (Fahrenheit v Centigrade) I find AGA easier to "get"


gennan wrote 3 months, 2 weeks ago:

I've been playing go for more than 30 years, including participating in many tournaments and I am not aware that territory scoring is a problem in practice.

The AGA rules really don't appeal to me very much. Having pass stones and a rule that white should pass last feels quite artificial to me. I understand how AGA rules "fixes" some issues, but to me it feels like they are a solution looking for a problem.

Yes, the official Japanese rules are not well worded, but with some adaptations, such as described by Robert Jasiek and other rule experts, territory score is fine, I feel.


antti wrote 3 months, 2 weeks ago:

The Tang-dynasty Chinese actually (apparently) used to have a rule that the players should play the same number of stones in a game. Back then that meant Black should make the last move, because White started, but now that would translate to White having to play last; this, in turn, makes territory scoring end up with the same result as area scoring.

While the above rule can also be considered ‘arbitrary’, it has a kind of elegance when you consider the principle of Yin and Yang, which the game is obviously grounded on.

Unfortunately, the above rule seems to break down when a situation such as the ‘sending two, returning one’ (https://senseis.xmp.net/?SendingTwoReturningOne) shows up: Black can continue sacrificing stones in such a shape, forcing White to play elsewhere until eventually White has to fill one of his groups’ remaining two eyes, killing it. Therefore, something like the superkō rule becomes necessary to ‘fix’ the game.

Source on ‘same number of stones’: https://www.usgo.org/sites/default/files/bh_library/historyofgorules.pdf (search for ‘equal stones’)


Events

All times are in Helsinki time (eet with summer time).


2nd Corona Cup on November 2 – December 15! Tournament details are here.


Gothenburg Online Open on December 5–6! Tournament details and registration are here.


New pairing for the online league made every 1st and 3rd Saturday of the month.


Public lectures on Twitch every 2nd and 4th Saturday of the month at 1 pm.


Jeff and Mikko stream on Twitch on Fridays at 6 pm.