Jigo implementation / bug not even assigned

I’m not really forgetting, because the komi being odd also sounds a bit arbitrary. I can’t imagine the game of Go being dysfunctional with area scoring, just because the komi was an even number.

Maybe there’s some argument to be made about various sekis and changing the komi? Can’t you just play with an agreed komi though no? Score it as it is?

1 Like

If there’s komi and the japanese rule is by default and most players (on western servers at least for what I know) use it, it’s simply from historical reason: Japanese players did a lot to spread go to the west.With my due respect ,a few decades later it’s just normal that players get some distance of this starting by the choice between different rules.

Of course this depends on if we are discussing “perfect komi” or “fair komi”. In the case of perfect komi, it is almost surely odd (assuming area scoring). Of course there’s a possibility that the perfect game involves seki and thus even komi. Either way, I don’t see the benefit of switching to an even board as black will tend to get more.

The komi has to be odd because: in Chinese counting the sum of Black points on the board and White points has to be 181, an odd number. Therefore the difference, which the score, also has to be odd. Therefore to bring the score back to 0 to account for White’s initial advantage, the komi has to be odd.

Also relevant: Please fix default NZ rules komi for ranked 9x9 and 13x13

2 Likes

If Black had 90 points and White had 90 points, and there was one dame from a seki, the score difference is 0 which is even?

You could say, “well that almost never happens”, but that’s also a bit of an arbitrary decision. Maybe you can base it on pro games, but it’s still a bit arbitrary.

What if there’s some board size, even or odd, or rectangular where the komi is exactly zero. What if it’s actually toroidal go? Or if it’s Go on a hexagonal/honeycomb board? Or maybe Thue-Morse Go (Thue-Morse (Fair Sharing) Sequence: A possible alternative to komi?)? I don’t expect all of the current pros to swap to that game because it’s in theory more perfect/fair etc. This feels like the current idea to swap to 7.0 komi.

One could say that maybe drawing with someone 10 times means you’re equally matched, but are you any less equally matched if you win 5 and lose 5?

You missed my qualifier : Chinese rules. The sum of points has to be 181 (excluding Seki of course). Most AI use Chinese rules because they are easier to implement (in fact I very much doubt if any AI use the actual official Japanese rules)

1 Like

I should have course said Chinese count, not Chinese rules, so my aside about Japanese rules was OT. I can never remember whether area counting versus territory refers to Chinese count or Japanese.

Games involving seki can potentially have an even point difference (even with Chinese). Unlikely, but it can happen.

1 Like

I believe the rating system on OGS already does handle draws properly. The only way to get a ranked tie is for the moderators to declare a tie by moderator decision. Currently, I think that calling the moderators for this manual fix (before ending the game) is the only work around if two players realize that a game should be scored as a draw.

Well, that’s a design problem to consider depending on the tournament structure. For tournament systems like round-robin, perhaps just splitting the point (1/2 - 1/2) makes sense. For elimination formats, I think a replayed game would be required. Note that for Japanese rules elimination tournaments, replaying a game is already a possible contingency.

Komi under Chinese rules is not typically odd nor even. Most often, komi is 7.5, which is neither.

Under area scoring rules, the score margin (difference between black and white’s scores before accounting for komi) is typically odd, since the odd number of points on the board are split between the two players. However, it is possible (although much less likely) for the margin to be even, if an odd number of neutral points are left unfilled (usually due to seki).

Hence, using an odd whole number for komi to allow jigo under area rules is typical, since it is believed that 7 is more likely to be fair than 6, just because it is unlikely that games end with an even score margin. Similar reasoning was used when the typical komi for Chinese rules games was switched from 5.5 to directly 7.5, since 6.5 would be less likely to make an impact.

There are 361 points on a 19x19 board.

3 Likes

duh. Sorry. 361. Still an odd number

1 Like

and you are being a little precious, you know what I mean about 7.5 being “odd”.

When the komi in Japan was 5.5 it was 5.5 in China. When they decided the komi was too small, Japan went to 6.5, China had to go to 7.5, because 6.5 gives the same results as 5.5 (again modulo seki as ever) and so would not have made any difference to game results.

1 Like

I understood your general intent, but it’s still a bit confusing to refer to 7.5 as “odd” or “even”, even though the reasoning is related to the odd number of points on the board.

Another way to think about 7.5 is to say “7 point komi, but white wins ties”, but it is also equivalent to say “8 point komi, but black wins ties”. So, it’s not natural to view 7.5 as either even or odd.

I would instead say “pedantic”

3 Likes

Right there in your flair!

Hmm so it seems setting a Komi of 0 in a handicap game results in an unranked game because it is “non-standard”. That’s a much more important bug than scoring jigo in NZ rules as a win by 0 for black. Waddaya mean nonstandard? That’s the way the game was played for a couple of thousand years and then along come these know-it-all upstarts with their newfangled electrickery who want to change the way the world works just because they think it might be too hard to program their toys (it isn’t, been there done that). I said it before, pernicious!

2 Likes

I’m not really of the opinion that that is the best way to handle it. Sure, it is somewhat rare to play with integer komi, and even more rare for jigo. But apparently there’s a notable section of the forums believing that integer komi is the ideal, and it’s quite the annoyance to then get things properly dealt with when it could be automated. I’d imagine most of the issue is the framework for catching a draw, which I’d imagine older code is not the most fond of considering how long this bug has existed. It’s not incredibly hard to implement to glicko-2 (as you just input s=0.5 for the game and you’re done), and bringing in moderators isn’t the ideal usage of time for a rather clear-cut case if it can be automated.

2 Likes

Sure you want to label yourself as a pedantic person?

I think they is no good reason not to have integer Komi. Even with it draws would be rare. There’s no downside to dealing with this properly as far as I can tell.

Pendant and proud!

3 Likes

Yes, I agree that it would be much better for it be simply supported than to waste the time of moderators to manually declare draws. My point was that the ratings system is no excuse for not supporting ties.

Do you really think that a self-confessed pedant would not know the meaning of the word?

It’s just a little bit of self-deprecation for the sake of humor and not taking myself too seriously.

7 Likes

I can imagine supporting ties in a tournament setting would pose the biggest issue. But yeah that’s not that huge and I could totally see a server supporting ties. (In fact, looking at the frontend code, it seems like the intention is for ties to be supported in some fashion)

3 Likes