My opponent refuse to remove the dead stones, and so, he wins. I refuse the score and remove the dead stones. He refuse the score, refuse to remove the dead stones… and so on, and so on.

Is there somewhere a referee?

Yes, sometimes the webmaster ends such a game. You should write to Richard.

I wouldn't call it cheating btw. Very likely your opponent doesn't understand the rules.

You can also keep playing some more moves and explictely kill his dead stones. i understand it doesn't work in tight cases, but often, in wildly won games against people thatdon't understand the concepts, it will work.

The case in question is tight unfortunately.

The game is not over. You have to play a2 now (and answer b3 with a3) in order to kill the b2 group. Otherwise b2 is alive in seki and white wins the game. White should have played a2 and won the game instead of arguing over dead groups. The typical play should have been: .., a2; c1, b1; pass, pass; and the result is w+9.5. Note that since neither of you can read the position, his behavior is probably not intentional.

Here is the game .

Also note that if you play a2 now, you win by 0.5, but if w had not played “142 m13” he would still be winning by 0.5 even if you killed b2. A good example why Chinese rules are so much better than Japanese rules for beginners.

Here are two questions I've asked before:

Is there any software that can score using Japanese rules without the human intervention that LG's current Go-scoring software requires?

Would Chinese rules allow the fully automated scoring of Go games?

To answer my own first question, according to the Rules of Go article on Wikipedia,

…the Japanese rules contain lengthy definitions of when groups are considered alive and when they are dead. In fact, these definitions do not cover every situation that may arise. Some difficult cases not entirely determined by the rules and existing precedent must be adjudicated by a go tribunal.

which rules out the possibility of software scoring every game.

Which is why top-level Go matches require 9-dan referees – it's silly, bordering on ridiculous.

The rules used for computer Go competitions are at the other end of the spectrum – a squeaky clean version of Chinese rules. No penalty for extra moves, so any dispute can be resolved just by playing it out. Those rules should be used for online venues, because of the easy software adjudication.

Now if the game is not over, why would there be a discussion here, with explicit instructions on how to play to not lose???

Wasn't that discussed ad noseum, 'DO NOT DISCUSS RUNNING GAMES'?

Yes, we definitely need Chinese rules. Japanes rules are a disgrace.

I agree with do not discuss running games. I made an exception because the players did not know it was a running game :-)

If LG were to change to Chinese rules, how would it work? Would the last moves of a game still be Pass, Pass, Score, Accept? Would a player whose opponent has just Scored be able to Accept or place a stone on the board, but not Score?

The change would fix a problem that I see in a forum topic maybe once or twice a year. Maybe scoring disagreements occur more often but aren't posted, or maybe I'm missing some because I don't always read the Go forum, but I have to admit it doesn't seem like a big problem. In any event, someone would have to ask the real Go players how they would feel about changing LG's scoring software.

On the plus side, according to that same Wikipedia article,

If the game ends with both players having passed the same number of times, then the score will be identical in territory and area scoring.

If that's correct, most games would have the same winner under either system. However, there was a post not too long ago about a bot that was using Chinese scoring internally and thought it had won a game by 0.5 points but LG's Japanese scoring gave it a 0.5-point loss. I don't recall which bot, let alone which game, but apparently there can be a difference. Or maybe the bot's programmer should have given it a different value for the komi to compensate for the difference in scoring methods.

Hmm… Why is it Pass, Pass, Score, Accept? Rather than a second consecutive Pass, the player whose opponent has just Passed may as well Score. I can see why we'd want to retain Score and Accept as separate actions, since the score may come as a surprise to the player who Passed and he may want to Place a stone rather than Accept.

I propose that if the scoring is changed to Chinese that the last moves of a game be …, Place, Pass, Score, Accept.

No, wait, why do we need Pass at all? Why not just Score? The last moves of a game could be …, Place, Score, Accept; a player could also Place a stone after his opponent has Scored (i.e. passed).

That actually sounds as a major insight! Cudos!

Oh, wait! Does zugzwang exist at all in go? If it does, then its not a good idea to make playing or accepting mandatory after someone played 'score'…

Or is “Pass, Pass” a way of keeping the Score button unavailable? If I had a Score button available, I could see who's ahead without having to look at the board and count anything (which I rarely do, which is one reason I'm such a patzer at Go). The “Pass, Pass” requirement means I need my opponent's cooperation to use the semi-automated LG scoring software.

> make playing or accepting mandatory after someone played 'score'

What alternative is there to Place or Accept after Score? With unambiguous scoring, a second consecutive Score is not an option. Resign?

Yes, there are differences between Japanese and Chinese scoring, but they are relatively minor, and one could argue that the Chinese rules are more logical even where differences exist. Thus, Chinese scoring gains a lot for a little. Adopting Chinese scoring world-wide would be sensible, but humans aren't sensible… ;-p

That said, now that Japanese rules are implemented, there are probably more worthwhile LG improvements for Richard to spend his valuable time on.

Interestingly, the game of “one-capture Go” is strategically similar to regular Go, even though the game is won as soon as an opponent's stone is captured. One-capture Go can also be played on arbitrary irregular grids (i.e. points having a varying number of neighbours), as discussed in Wayne Schmittberger's “New Rules for Classic Games”.

[ https://en.wikipedia.org/wiki/Capture_Go ]

[ http://www.amazon.com/Rules-Classic-Games-Wayne-Schmittberger/dp/product-description/0471536210 ]

Note that the LG forum automatically hyphenates any word containing “skript”, so “description” in the amazon URL above shouldn't have a hyphen.

Next time I'm in a very close match, I'm gonna have to remember the old “my opponent's cheating” line so that I can get your advice on what moves I should play next.

@MarleyGhost - real good question. BTW, I must clarify that I'm not good at all in go but my brother is quite a keen player. I can recall at least one occasion in an OTB games, that me and my equally patzer opponent both passing when bro came up and said “game's not finished, u 2 must decide what happens here and there…“. How would Chinese counting rules deal with such an 'unfinished' game?

Furthermore, thinking on debugging techniques, how would counting software deal with an absurdly early attempt to count score? White automatically wins by komi?

With Chinese rules, either the players agree on the outcome, or they play on until it is fully resolved. No biggie, since it can't change the result (apart from blunders).

With Japanese rules, as i understand it, a committee of one or more referees who are as strong or stronger than the players convene, and come up with some arbitrary ruling based on their opinion on that particular day. This might become precedent for amending the rules of the game (as with the example of “bent four in the corner” (when was that? a thousand years ago?)). I can't begin to express how abhorrent that is to me, on so many levels… :-/

With a software referee for the simple Ing GOE rules, the regions may need to be fully closed off first; otherwise the result might be “don't know”, or might be wrong, depending on the implementation. [Note: i've never been an arbiter for GOE competitions, so i'm not 100% sure on this.]

Here is a pdf with the complete Ing rules:

[ http://www.usgo.org/resources/Ing%20Rules%202006.pdf ]

Related links from the usgo.org source:

Discussion of Ing “Stones and Spaces are Territory” (SST) rules:

[ http://www.usgo.org/resources/KSS.html ]

Ing rules, 1991:

[ http://www.usgo.org/resources/SST.html ]

Counting example:

[ http://www.usgo.org/tournaments/Resources/IngCounting.pdf ]

There's been extensive discussion about this here in the past. I'm no expert, but I've learned that Chinese is not the only area-scoring and Japanese not the only territory scoring. I forget who, but someone made a good case that the AGA (American Go Association) version of area scoring was ideal. for here.

To FatPhil and others. The point is not to know what to play next. We both (my opponent and me) agree to determine the score, even if the game was not finish. We din't noticed that because our level is pretty low. The point is: which are the stones to remove? Who wins, AT THIS POINT?

The outcome of the game depends on whether b2 is dead or alive (in seki).

At this point (b to move), b can play a2 and kill it winning the game. With w to move, w can play a2 to make it live and win the game. With Chinese rules, white wins anyway: If b kills b2, it is w+1.5 and if w makes b2 live it is w+12.5.

I can understand that playing a2 now is a bit lame since it was pointed out in the forum, so resign is also an option.

To be clear, usgo.org is the AGA.

Ing put *a lot* of thought into his rules (refining them over years), and used them for the Ing Cup tournaments that he sponsored. It looks like the AGA has adopted them, or some other formulation of Chinese rules (although they do not state that).

The Tromp-Taylor rules are even simpler (John Tromp and Bill Taylor are both mathematicians). I don't know if they've gained wide-spread approval though (the world is irrational:).

Links to the Tromp-Taylor rules:

http://homepages.cwi.nl/~tromp/go.html [ or http://www.cwi.nl/~tromp/go.html ]

http://en.wikibooks.org/wiki/Computer_Go/Tromp-Taylor_Rules

Some useful comparisons or interesting discussions:

https://en.wikipedia.org/wiki/Rules_of_Go

http://www.britgo.org/rules/compare.html

http://www.cs.cmu.edu/~wjh/go/rules/

http://www.cs.cmu.edu/~wjh/go/rules/simplest.html

http://senseis.xmp.net/?TrompTaylorRules

There has been extensive discussion of this topic in many places, not just LG, so there doesn't seem to be much point in re-hashing it here again.

I looked at the Tromp-Taylor rules. Rule “9. A player's score is the number of points of her color, plus the number of empty points that reach only her color.” is where scoring software comes in. I admit, rule 9 can be implemented as unambiguous scoring software. But implementing rule 9 as written means that if player X has even a single stone in what is in fact player Y's territory, that territory does not count toward player Y's score. I'm sure the players would rather not have to play to the point where all of player Y's territory is free of player X's stones.

We could write software that implements the spirit of rule 9 with some additional algorithms to recognize some X stones are dead and apply rule 9 after removing those dead stones. But there are limits to how well software can distinguish live stones from dead stones. So the most the scoring software can be guaranteed to do is come up with a lower bound and an upper bound to the score.

If those bounds do not coincide, and in particular if the bounds don't both name the same player as winner, there can still be a disputed score. What a revised system could do is guarantee there aren't two consecutive Score moves (let alone 94).

They also give a procedure for removing dead stones and arriving at agreement on the outcome of the game. The commentary section puts things in a way more familiar to Go players, which they kept that separate from the succinct formal rules.

" 8. As a practical shortcut, the following amendment allows dead stone removal:

After only 2 consecutive passes, the players may end the game by _agreeing on which points to empty_. After 4 consecutive passes, the game ends as is.”

Here's the link on the agreement protocol: [ http://homepages.cwi.nl/~tromp/agree.html ]

So, all the hard work on cleaning up the protocol has been done (a few times, in fact). It's just a matter of implementing those more precise methods.

There are probably several software implementations available (to use as a precise specification and for testing, if nothing else). Tromp gives one written in Haskell, and there may be other programs for Chinese rules or New Zealand rules.

IIRC, Fuego handles both Chinese and Japanese rules, to some reasonable degree of completeness and accuracy. Fuego is the University of Alberta program that has won a few recent championships, and is now open source (C++). Martin Mueller has been writing Go programs for decades, is an amateur 5-dan or 6-dan player, and has an intimate knowledge of combinatorial game theory (e.g. cgt temperature applied to Go endgames), so Fuego's understanding is probably pretty good.

Thanks to you “erratic”. To close this neverending game, I resign with no regrets.

Have a good day!