Meta-standards in Go that irk you

General Go Discussion
1

As much as I love Go, there are several widely held meta-standards around its play that I believe limit its growth and appeal. I wanted to share my thoughts on these issues and see how others feel about them, as well as if I’ve missed any points:

  1. Small Board Beginner Stigma: Modern trends in gaming, along with other aspects of modern life, favor short, to-the-point games that can be played on a small phone screen. A 9x9 Go board is slightly less complex than chess, and a 13x13 board is already much more complex than chess. These board sizes can be comfortably played in less than 10 minutes, unlike the 19x19 board, where 10 minutes feels rushed for most players. The insistence on making 19x19 the standard (rather than an option) likely limits Go’s potential as a game that can be easily and casually played. I think OGS is ahead of the curve by having blended board ratings and small board tournaments, but a bigger push needs to come from the major Go associations in Japan, Korea, and China.

  2. Hegemony of Territory (or “Japanese”) Scoring Outside China: Territory scoring is more convoluted to teach absolute beginners, as it may discourage playing out moves to determine life/death due to the penalty of playing inside your own territory. There are no disadvantages to area scoring—some argue it’s slower for physical counting, but this is demonstrably false if you’ve played physically in China. Others claim territory scoring is more ‘granular’ or ‘skill-based,’ but there’s no empirical evidence that it actually yields more skilled players or better play.

  3. Non-Fischer Time Controls: While some people are uncomfortable with it, Fischer time controls are objectively better compared to the byo-yomi and Canadian byo-yomi systems that dominate most servers. OGS is ahead of the curve by promoting Fischer time controls. The ship seems to be moving on this one, although agonizingly slowly.

  4. Micro-Handicaps: Rating differences correspond to expected win probability, not expected score/territory, especially at higher ratings. Many servers (including OGS) implement micro-handicaps for minor rating differences. I believe a win probability split of 40/60 or even 30/70 is still enough to warrant an even game (I personally would accept a 10% win probability without a handicap). These micro-handicaps likely add destabilizing noise to the rating system.

  5. Insistence on the Kyu/Dan Rating System vs. Numeric Ranking Systems (ELO, MMR, Glicko): Many major servers use outdated ranking algorithms. Standards for nominal ranks vary greatly by association and server, making them almost meaningless unless you encounter another player from the same server/association in real life. The supposed advantage of this system—that a one-rank difference corresponds to a one-stone difference—is flawed (see point #4). OGS is ahead of the curve by using a Glicko2 system.

  6. Existence of Ranked Handicap Games and Handicap ‘Tournaments’: Allowing ranked handicap games can add noise to server rating systems, as rank differences correspond to win probability differences, not score differences. For smaller servers like OGS, this may be a necessary evil, but it shouldn’t exist for larger servers. Handicap ‘tournaments’ (such as some held by the AGA) are an oxymoron akin to a participation trophy, as tournaments are meant to be a competition of skill.

  7. Aversion to Ties in Ranked Online Games: The standard remains for ranked online games to have komis ending in .5 to avoid ties. However, evidence suggests that integer komis are more balanced (see KataGo). I think it’s better to be fair and call a tie a tie rather than a win for white.

These are my peeves/hot takes, but I’m interested to hear if others think I missed something or disagree with me.

14 Likes
2

A few thoughts

I agree that territory scoring is considerably less elegant. Note however that, while Japanese rules remain dominant online in the West, tournaments are played under AGA or other western rules.

Quite relevant given the recent discussions in making Fischer the default on OGS!

Probably my fault but I have no idea what this is referring to.

I don’t really see the issue here. The Kyu/Dan system is simply a different way of displaying your rank (and has nothing to do with the question of how rank adjustments are calculated). The fact that it is supposed to map a one-stone difference is a rather elegant anchor, and easy to understand (though not-so-easy to achieve of course).

Displaying a number instead would have no impact on the discrepancy between servers and associations; comparisons would still be meaningless. What you seem to be advocating for is a unified ranking system, which would be a great thing indeed, but I’m not sure how that relates to the Kyu/Dan system.

Also don’t understand your point here, as tournament results also depend on whether you win, not the score. Sorry if I’m missing something.

Agreed, but it’s not limited to online games. Komis are defined by the rulesets, and online servers simply refer to such existing rulesets.

3 Likes
3

There’s a certain beauty in the strategy involving 19x19 games that you cannot find in 9x9 or 13x13 games.

Also, I’d rather like to push back on the recent idea that everything has to be bite-sized in modern life.

To comment on this a bit more, Glicko or Elo (not entirely capitalised, it’s a personal name after all) systems are not comparable between different pools. That is, the Glicko rating of someone playing Go on OGS cannot be compared to the Glicko rating of someone playing Go on another server or association.

Namely, consider an Elo system used within a server like OGS, where most players are amateurs, and one that is used to rate professional players in official tournaments. Both systems will have the same average rating for their players, but clearly one pool is far stronger than the other.

That’s why it’s a bit silly when someone says “I’m this-and-that Elo” without specifying within which system.

There’s other problems with such systems, namely that they assume the player base is a normal distribution, which it generally isn’t (intuitively, the weakest players are less weak than the strongest players are strong).

That said, OGS shifted the kyu / dan ranks to make them align as close as possible to the AGA and EGF ranks, so there is some comparison possible in that regard, ironically with the kyu / dan system of OGS.

I don’t see the problem here with larger boards. What’s wrong with having a game having a winner? I personally find ties boring. Look at what they do for chess!

4 Likes
4

What’s wrong with a tie if both players played equally well, though? Especially if AI confirms that this result would be more fair.

The issue in the Chess worlds comes from ties being excessively common (in large part due to players being able to strategically aim for a tie instead of a loss). This isn’t possible in Go, ties would remain a rare event.

4 Likes
5

But only more fair between players who are really the same strength. If 1 rank difference is already considered to be around 6 komi, then it’s very rare you even play against an opponent with the same strength anyways.

Not in professional games

2 Likes
6

Obviously more common than in DDK games, but the amount of 0.5 wins is still pretty rare. If your disagreement is on the meaning of “rare”, at least let’s agree it would never be an issue comparable to Chess.

4 Likes
7

I do agree on that. :slight_smile:

But still, a tie feels a bit boring to me.

2 Likes
8

That strikes me as about right, but I’m biased towards Go, lol

I think this point is overstated. Yes, 19x19 is the standard, I don’t think we should shy away from that, but there is also a devoted 9x9 community here, and 9x9, 13x13, and 19x19 are all considered “standard” sizes in a looser sense. All three are standard, but 19x19 is more standard, so to speak. I think it’s in a good place; I hope more people will check out 13x13 (I think it’s underrated), but I don’t see any benefit in forcing it

I want to eliminate the blended board and time control ratings, promoting the individual board size and time control ratings to first class citizens, analogously to what lichess does. Far from encouraging small board play, the blended ratings shunt it to alt accounts where it won’t impact your main account rating. Eliminating the blended ratings would on the other hand would promote taking all three sizes seriously, as they new have ratings of equal prominence to your 19x19 rating

I do love the ASTs on a variety of sizes

emphasis on “may”. It does make playing out L&D disputes wrong in all territory scoring rules implemented online to my knowledge, but it does not have to (cf Lasker-Maas Rules

It is objectively more granular (unless you meddle with Area Scoring with a button), but I like draws as a slight possibility, so I rather like the slightly less granular nature of Area Scoring (as well as the nice coincidence that perfect komi is an odd integer in Area Scoring)

While I like Fischer (I use it pretty much exclusively), and I agree that while Canadian Byo-yomi is a reasonable compromise solution when stuck with analog clocks it cannot compete with either Fischer or Byo-yomi in terms of goodness, I think Fischer and Byo-yomi both have pros and cons

  • Fischer
    • Simple (main time, +x seconds after every move (optional™ cap))
    • Flexible (can play faster on easy moves, and spend the extra time on harder moves, and gives you a better chance to avoid losing on time due to an internet blip)
    • Bloated Bank (in the endgame especially, higher increments can result in accruing significantly more time in your bank than you actually need in order to find moves. Since you do have the right to use your time even if it’s just to twiddle your thumbs, this is awkward, whereas Byo-yomi prevents unused time from piling up never to be (non-stallingly) used)
  • Byo-yomi
    • Simple (main time, then x seconds per move (optionally multiple periods)
    • Consistent (keeps the moves coming at regular intervals once players enter Byo-yomi, which is good for spectators)
    • End of Countdown Moving (once byo-yomi starts, one is generally best advised to use their whole period every move, thinking about other moves during the extra time if the current move is trivial (denying the opponent thinking time occasionally makes this not so clear cut in certain situations, but in general it holds), whereas Fischer allows you to simply play the easier moves fast, and use that time when you actually want to use it)

Which you prefer will largely come down to how you weight these factors (and probably others I missed off the top of my head). My solution is to use Fischer with no cap™ (OGS doesn’t allow true no cap, but it allows caps high enough that they may as well be no cap), but largely fix the biggest flaw in Fischer (Bloated Bank) by using a short increment of +10 seconds per move. I generally play with 20 minutes main time, which gives a pretty slow and serious game, but not too slow, but the overall speed of the game can be easily adjusted up or down just via changing the main time while keeping the +10 seconds per move increment that I feel works well to cover for Fischer’s biggest weakness

OGS attempts to empirically tie the glicko-2 ratings (which do indeed correspond to expected win probability, not expected necessary reverse komi and/or free moves at the start of the game to bring the win probability to 50-50) to the kyu-dan ranks in such a way that the latter do correspond to the expected necessary reverse komi and/or free moves at the start of the game. There are some subtle flaws in the current methodology, but those will be fixed in the v6 ratings update

Far from adding noise to the system, handicap games are integral to how OGS gets the necessary data to empirically determine how best to correspond glicko-2 with kyu-dan. Many people enjoy handicap tournaments as weaker players have a chance to win, and stronger players must play their best in all games

Personally, I rarely play handicap games, but I would play more if automatic reverse komi handicap was an option, as may be the case in the v6 ratings update

I agree here; one of the major reasons I use NZD rules is that it uses 7.0 komi which (in area scoring, which NZD is) is both fair (gives the best chance of a draw) and perfect (gives the closest to 50-50 win probability)

Note that perfect komi for territory scoring (without button shenanigans) is 0.5 points less at 6.5, and fair komi in territory scoring is 6.0. One advantage of Area Scoring is that fair komi and perfect komi happily coincide

3 Likes
9

Agreed. I think 19x19 is the standard of standards for good reason, but I don’t think that contradicts valuing 9x9 and 13x13 for their own unique feelings

I also very much agree that just because something can be bite sized does not mean it should be

Some people dislike ties, other people don’t mind them. Chess has a far bigger draw problem than Go will ever have. I like when games have a small possibility of a draw, but Go will never have to worry about draws being anywhere near as common as in chess, which I think has too many draws

Yep. The only issue would be if there were way too many, like in chess, but that won’t happen for Go as Go is much more granular than chess

100%

I think this was a typo; 1 rank difference is equal to twice perfect komi (that is, 13 points for Territory Scoring and 14 points for Area Scoring)

Agreed, @qnpnpmqppnp

1 Like
10

There’s a balance between depth and approachability for general public. There’s high art and pop art. I don’t think that everything and everyone should lower themselves to the lowest level just to appeal to the broadest audience. You can think of go as SSBM of board games. There’s certain auditory for such kinds of game.
Just go somewhere else an enjoy other thing, no need to destroy something others love.

That’s just not true. There’re Korean, AGA and other rulesets that popular in their respective areas. Outcome is the almost same and should be ignored by most beginners.

Please elaborate on your claim. It’s not obvious to me what you imply by objectively better.

I don’t understand what are you talking here about exactly.

When in Rome do as the Romans do.
I see no need to reinvent the bike. Amateur kuy-dan ranks are built around 1 handicap stone strength difference in mind. Isn’t it elegant enough at this point?

What’s up with other options existing?

I think you’re just missing the point of handicap games. It’s meant to equalise probabilities of winning or loosing for both players regardless of skill difference based on inherent qualities of kyu-dan ranking system.
If you don’t want or like or can’t tolerate weaker player having higher win probability or stronger player- lower probability, then don’t use it. Again, there’s no such need to outlaw something that does not corresponds with your own views.

No aversion on my part at least, I just think it’s more satisfying. Only aversion that I can hear of is coming from you not being able to tolerate fractions komi. I don’t see how this limits appeal or growth of the game.

3 Likes
11

Among 7 arguments you made I only see 1st argument having any relation to restricting go growth and appeal to the broader audience. I think it could be more broadly summed up that go is just a hard game which requires a lot of commitment to even begin to be even decent at.

6 Likes
12

While I agree with a lot of your preferences (Fischer, Area Scoring), the only point I think plays significantly into the growth and appeal of Go is probably the smaller board size, and maybe the scoring system although it’s debatable whether one is actually easier to learn than the other.

You don’t need a superhuman AI to know that perfect komi is integer komi. The decision to use 0.5 is pragmatic one as there are certain contexts where ties are unacceptable. This is not unique to Go, see Wikipedia: Tie-Breaker

That said, I do enjoy a “tie-able” game in more casual settings :slight_smile:

I disagree strongly. If anything, handicap enables Go to simulate a larger player base than it has, which is important, given network effects.

One of mine (and I know it’s unpopular) is the fetisization of Japanese terms. Like fine if there’s no English equivalent, use Japanese. But often there are direct and simple translation, but we use the Japanese term because it’s mystical and exotic :crazy_face:

3 Likes
13
  1. Long and pointless forum threads that don’t achieve anything.
    :smiling_imp:

(maybe that’s just the internet in general)

2 Likes
14
1 Like
15

The first time I found out about chess having a lot of ties was through the World Chess championship. The two players tied both in the classical and rapid and in the end the winner had to be decided through blitz. They also had to come up with something called Armageddon just to decide the winner. This makes it boring for the spectator because if you are going to decide the winner through the special method, what was the purpose of all the games that have been played? Why not just go to the special method straight away?

It’s like watching close to 2 hours of a football game only to have it end in a 0-0 draw. In the end they have to decide the winner through PK which is not really reflective of their full skills. Then why not just have the PK straight away. (A bit exaggerating but yeah you get the point)

So I’m really glad that this is not happening in Go and a draw only happens in very rare cases.

6 Likes
16

Fischer time is great for correspondence, where plenty of time is available. However, in live games it requires greater time-management skill than byo-yomi, which makes it more difficult for weaker players IMHO. Because the available time changes with one’s every move in Fischer, keeping track of one’s time is a constant distraction; byo-yomi has no similar distraction.

The OP failed to define “micro-handicaps,” which is why several of us have no idea what he is talking about in that paragraph.

2 Likes
17

My proposal to fix Soccer:

  1. Remove the Offsides Rule: if you’ve got the teamwork to get the ball past the opposing defense, you’ve earned the shot on goal
  2. Remove the Goalie: In order to reduce tie games and better discriminate team strength over the course of a game, you need to increase the number of goals scored throughout the game, and to do that you need to nerf defense
  3. Replace the current tiebreak method with a sudden death overtime in which every n minutes (value of n to be determined by strong players who can make good judgements about the best value) the number of players each team can field reduces by 1. If no goals have been scored by the time it gets down to 1v1, it continues until a goal is scored
3 Likes
18

Agreed; I like draws, but only if they’re not too common

2 Likes
19

One thing that slightly irks me is the use of a clock in casual non-competitive games. I understand that the clock is mandatory for scheduled tournament games, but if i’m just playing with a friend for fun i prefer to play without a clock. I don’t feel any joy for winning a game because the other player ran out of time :<

Yeah, its a pity that 9x9 and 13x13 are always just “side tournaments” during big go events, while the “main tournament” is always 19x19. Would be great to have proper RL tournaments on some other board sizes too. I personally enjoy 17x17 and 21x21 boards too, but theres never any tournaments for those :<

One big one: you cannot redo the count if theres some disagreement. With territory scoring you can re-count the territory as long as the stones are on the board, but if you do area scoring and then disagree the amount of territory someone had, you cannot simply count it again without replaying the entire game. (Tho this only applies to RL games on physical board)

I fully agree on fischer, i think its by far the best time control for go.

I think the main reason is that while numerical rating systems are “a must” in digital era, people still prefer using the old kyu/dan system just because its a tradition. Combining the numerical values with kyu/dan system seems to be a struggle for all servers and national associations… xD

With this one i disagree a little bit. I get the point that tournament games should be between players who have roughly equal skills, but sadly thats not usually possible without tens of players. I think the main reason for using handicaps is providing a challenge for the stronger player. Go is at its best when its hard, so it makes sense to use handicaps when theres a big gap between players skills. Its no fun to just outright crush the other player and win without any effort >__>

I fully agree, unless the game is part of elimination tournament then there is no good reason to artificially avoid ties. I’ve been in some RL macmahon tournaments which used integral komi and never had any issues with games ending as ties.
OGS can also handle ties perfectly fine, tho only the New Zealand rules have integral komi as the default. (you can also use any other rules and manually set the komi as 7 or 0 or whatever, but games with custom komi will always be unranked on ogs)

2 Likes
20

Its fun sometimes. Its exhausting to always have 50% chance games.

5 Likes