Seki?

How do you determine seki during scoring? It didn't make a difference for the results in my game, but what would I do to score it?

http://www.littlegolem.net/jsp/game/game.jsp?gid=1205987

when you click 'score' it would make the area under question red, and not count it for anyone. Is that not what you want?

I'm not sure which scoring rules we really use here, but I'd think the seki area (the two points in the lower left) counts nothing for nobody?

using the japanese-like rules (but not really japanese...) used here, there are no points indeed in the seki.

But the scoring algorithm, as can be seen in this very game, does score the eyes d1 and a1 in seki. In this game, it doesn't matter because there is one such point for each side. But there are seki positions where one side has such a false point and the other doesn't.

As antony pointed out, "scoring is japanese-like, but not really japanese."

The difference is that seki eyes count as points here. ( Similar to IGS )

The example here is not complex enough to show a problem (both have one additional non-japanense point), but see http://senseis.xmp.net/?SekiWithEyesQuestion1

If this one point does make a difference in the result, you can either ask your opponent to resign, or you accept the fact that scoring here is not japanese, but IGS style.

Resign does not fit to the Hahn counting system, however.

To get a "correct" score you could agree with the opponent to place stones inside your own territory to compensate for the extra seki stones.

Anyway, the game won't end until somebody clicks "Accept" or times out ;)

Is there a way to get the scoring system to recognize that it's seki, even if the points inside are counted? I couldn't figure it out in the game I referenced. One of us had to be marked dead. As you can see from the score, the system reads that black group as alive, and the white stones inside as dead.

Dodnt rely care because I won anyway, but how would I do it other than begging for a resignation if it does indeed affect the outcome?

I did t

Erp sorry for formatting. iPhone got goofy after the first paragraph.

But in other words, the area wasn't marked red at all, and I couldn't figure out how to make it such.

The scoring system here does not recognize even the concept of seki. So there's no way to get it to recognize a particular area as seki. All it knows are captives, live stones, dame, territory, and komi.

Gotcha, thanks!

On a side note, it's not just two points. With the territory and captures, it's seven (or six, depending on the scoring system). Or maybe I missed what y'all were saying.

seki is mutual life, so don't mark the stones as dead, and the scoring system will recognize seki.

not that easy, xed_over, as you can understand from reading the thread. under japanese rules, if the system recognizes seki, the eyes enclosed should not be counted as points. But here they do.

It is rumored that go can be played without a computer although many of us never did. I couldn't find opponents in these latitudes :-)

What LG implements are Computer Japanese rules. The only Japanese rules I have ever used. They are Ok. A point surrounded by living groups of the same color is a point. Why is that wrong?. If the living groups are of different color, it is dame. Dead groups count as the opponents. It makes perfect sense to me.

Do you have any reference of what you call "Computer Japanese Rules"?

I have played tons of go games using my computer. some of them were against an AI and the others were against real people around the world.

I've never seen nothing like "Computer Japanese Rules". Not in KGS, not in IGS, not in DGS, and not in any software I used when I was younger.

Under Japanese Rules, the points surrounded by groups in seki are not counted. However, LG has "very partial implementation of the japanese rules".

On a slightly unrelated note, where do you live? I see a spanish flag in your profile. Are you living in Spain? If so, you have plenty of players to play with here in Spain ;-)

Hmm, at the time I seemed to feel that the system was automatically declaring one of them dead, but now I'm wondering why it would do that? In current games I have to mark what's dead. Was there a software update, or am I just a screwball? Anybody care to replay this game with me to see what happens? I'll even do the trouble of letting you know your next move.

No, in this game Black marked your two stones as dead. I wonder why you accepted :-)

I seem to recall trying to rescore it, and not being able to have the area be zero. It was either White's or Black's, but not Red's. I accepted because I won the game anyway, so it didn't matter.

invite me to a 9x9 game, and we can recreate the situation. You can rescore it to be seki :)

Cool, thanks! Your first move is F7.

Ah, yes, I was just a bit confused, misunderstood what was being said. Yeah, I just accepted them as dead because it made no difference to the result of the game.

as expected, he was able to set the marked-dead back alive and thus the dame be red. However, as we pointed out, points in seki are marked as points, but shouldn't. I don't really know why they are not points under Japanese ruleset.

In case you are wondering, usually Black has as many points in seki as White do, but you can create a seki in which black has no points and White 30 or 40 points :-) so there's indeed a real difference :o)

Thanks for testing it with me!

Hi Gregorio:

I am a modest go programmer and all programs I have seen the source code: my own, gnugo and fuego count the score with dead or alive groups. There is no such thing as "alive in seki". KGS has some "smarts", maybe it guesses somehow, I don't know. How do you tell KGS that a group is alive in seki? You mark it dead or you don't. Players sometimes add stones to compensate the difference with human rules when it makes a difference. Usually MCTS programs play Chinese rules to the bitter end and kill every thing but counting the number of passes to keep the difference between Chinese and Japanese. Counting the cells surrounded in seki as zero is very complicated and I haven't seen it done, but maybe some commercial programs like Zen of MFOG do it. I don't know.

BTW: I live in Tenerife, but I am from the Basque Country.

I just loaded the game in question into CGoban3, and it recognized it as seki, with no points for either player. So KGS figures it out somehow.

an unfillable dame is not that hard to detect, imho :)

Modern engines use Monte-Carlo Tree Search (MCTS). Seki is detected in the tree. This means, you don't play the self atari move because it makes you lose. But the board is counted at each playout. The number of playouts per second is the holy grail of MCTS so you cannot detect eyes with Benson's algorithm or things like that at the end of the playout. You just count the points. Classic engines like gnugo, which are not strong today, use Benson's algorithm so they could understand that a living eyeless group lives in seki, but they don't. It counts as just living. At this moment a tournament is being played on KGS (Computer Go room) where MFOG is winning doing 9 million playouts per move. At that rate, if it takes too many clock cycles it is not worth the pain. What is and what is not easy depends on how much it costs. The question is, is it important? It is an arbitrary and annoying rule, but I agree that if it was a priority solutions could be found to work around it. Maybe for a commercial engine it is, because humans are used to count it the complicated way.

it's not a matter of counting it the complicated way. A game can differ by 20 points if you count eyes in seki. Of course you can overrid that rule and then you are playing a different game. If cgoban3 detects seki, then it's doable.

And I'm not talking about AIs playing a game. All this thread is about setting the status of a group in the end and counting the points under japanese scoring.

Why is it wrong that a cell surrounded by living groups of the same color counts as a point? It makes 100% sense and many rulesets accept that.

> A game can differ by 20 points

No it almost never makes a difference.

a. There need to be groups in seki.

b. Some of these groups must have an eye.

c. The number of eyes must be different for both players.

d. That difference must decide the game.

One game in very very many.