OGS rank mixed with all board sizes?

I agree that operating ranked-bot-tournies makes no sense.

Do we really do that? Dang it if we do :slight_smile:

However, I don’t really agree with the implication that this would have a signifcant impact on OGS ratings in general, because those tournies are a small small subset of players.

What are we defending against?

An assertion that “OGS ranks are broken” that is not backed by any data, it seems.

I think that if you come with assertion that “the team” - of volunteers who are passionate enough about the site to help run it - have “got it wrong” then you surely expect them to

  1. defend what they have done - they wouldn’t have done it if they didn’t think it was right
  2. expect a decent argument from you about why it is wrong.

One thing I personally like about the community here, and take pride in doing my best at, is admitting I’m wrong and changing when it’s demonstrated that I’m wrong.

But that doesn’t mean you get to say “its broken” without a pretty convincing argument…

5 Likes

Isn’t this self evident? Wouldn’t it be pretty weird if the volunteer moderation team didn’t participate much in discussions and didn’t think favourably of the site and its features more often than not? None of us think it’s perfect here, but we all think it’s pretty awesome :slightly_smiling_face:

7 Likes

It predates superhuman AI.

It’s another of those things that existed on old OGS and carried over to new OGS without sufficient thought.

Like OGS-critical moderators such as myself :wink:

3 Likes

There’re a lot of fake ranks, like foxy ranks are ridiculous, people still use them just fine. I mean internet ranks are fake in general.

5 Likes

in order for a server to be good, you want their rank to represent their strength as close as possible, everything I have brought up so far is potentially loopholes where people can exploit their ranks

just saying you know the server has done a tremendous good job if someday people can use your servers rank to determine whether you’re good or not

the irl rank is the most authentic one since 1. won’t worry about running into cheaters 2. everything is done face to face, fair and square and etc etc. in terms of online ranking, there’s a lot which we can’t control but what we can control we need to make sure to make it as best as possible

ranks were separate on board sizes before on OGS and 19x19 rank was very outdated for many people
both systems has problems - old and current
but current system looks like has less problems than old
correlation between 9x9 and 19x19 strength exists

5 Likes

I don’t even understand what’s the point of making a separate rank for 9x9 and 13x13,since no one has any ranks in either 9x9 or 13x13,those who spent the longest time on those board sizes would be regarded as pros in the world of smaller board sizes. the rank itself became meaningless since no one really have any training or practice. people usually play smaller board sizes either to gain a better understanding of the game or playing for fun

You keep saying our ranks are meaningless without offering a shred of evidence.

I understand that you have pointed out potential ways in which OGS ranks might be undermined.

Every system has potential flaws, and we do work to correct those at OGS.

However, just pointing out that “Hey OGS has some ranked events with bots” is far far from establishing that this causes any problem whatsoever.

Meanwhile, someone actually did measurements and found that using combined ranks predicts match outcomes better than separate. This is the ultimate evidence for a ranking system working.

Yet you appear to be ignoring this fact.

snakesss pointed out that some flaws might have been found in that analysis. Great - let’s explore those. Come up with corrected analysis. This is how you establish that there’s a problem, not putting forwards possible scenarios that might have an effect.

I tend to agree with this. Since these “ranks” are not used for anything, it’s possible that presenting them just creates confusion. On the other hand, IIRC folk objected strenuously to the idea of them going away.

1 Like

Why do you assume that the number of intersections suddenly makes it nonsensical to try to assign a ranking to the game? I’d get why ranking in 2x2 or 3x3 Go may not make sense, but 9x9 is perfectly valid and I wish you good luck finding someone who is a 20k in 19x19 who can beat a professional Go player on a 9x9 board.

7 Likes

I used to believe that too. Intuitively it makes a lot of sense to believe that ranks should be kept separate between 9x9, 13x13 and 19x19. Thank you for noticing this and thank you for making a suggestion to improve OGS.

Also, it makes a lot of sense to believe that ranks should be kept separate between lightning games, non-lightning real-time games, and correspondence games.

Note that the resulting nine separate ranks can be found on a user’s profile. For instance, on this page you can see my ranks, separated by 9x9, 13x13, 19x19, lightning, non-lightning real-time, and correspondence: ArsenLapin1
Screenshot from 2022-05-26 16-24-05

So why is OGS not using only the 19x19 rank? Why is OGS using 9x9 and 13x13 games as well when computing the rank? First, let’s examine your arguments.

When I started playing on OGS, I played 97 games on 9x9 against a bot, and lost 95 of those games. Note that I was quite proud when I finally managed to win 2 games. Now, you might think that my rank would never recover from those 95 defeats. Well, it turns out that I’m now 2.9 dan OGS, which happens to be almost-exactly my rank in real life at my national association. I believe this has been explained by Sofiam:

So, I was allowed to rapidly recover from all my losses on 9x9.

Let’s move on to your next argument:

Yes? I think it does. Do you have an example of a top pro losing on 9x9 against someone who isn’t a top pro? Do not underestimate top pros. They have played the game of go a lot. That includes 19x19 as well as 9x9 and 13x13. Top pros are strong. If you believe otherwise, my guess is that you are severely underestimating top pros.

I’ve been participating in 9x9 and 13x13 tournaments in real life. There are some, in several occasions. Usually these tournaments are played without handicap. The ranking system from our national federation only takes 19x19 games into account. But guess what: there are no surprises during the 9x9 and 13x13 tournament. High-dan players consistently defeat low-dan players. Low-dan players consistently defeat low-kyu players. Low-kyu players consistently defeat high-kyu players. In other words: the 19x19 ranks recoup perfectly the 9x9 and 13x13 results.

As it turns out, the game of go remains the game of go, no matter the size of the board.

For a whole year from may 2018 to may 2019, I counted the games I played. I played over 1500 games in one year. More than half of those games were 9x9 games. And I didn’t have any difficulty finding opponents, which tends to show that other players are playing on 9x9 too. Perhaps 9x9 and 13x13 are uncommon for you, but they’re not uncommon for everyone. In the past 4 years I think I’ve played at least twice as many games on 9x9 than on 19x19. Note that 9x9 has only 81 intersections, whereas 19x19 has 361 intersections: as a result, 9x9 games eat a lot less time than 19x19 games, and so it is easier to play many 9x9 games than to play many 19x19 games.

Sorry, what? What do you mean “supposed to”? On whose authority are we supposed to play on 19x19 and not on 9x9 and 13x13?
Every 19x19 board at my local club has a 13x13 on its other side. We also have a bunch of 9x9 boards. When I introduce beginners to the game I always teach them on 9x9 and encourage them to play on 9x9 first. Then on 13x13. Then much later, if they want to, they play on 19x19. Go works really well on 9x9 and 13x13.

I started this post by saying that it was natural to believe that ranks should be kept separate between different board sizes and different time settings. However, OGS is doing a little better than just believing. Anoek posted a long and detailed analysis about the ranks. The conclusion of that analysis was that the ranks were a lot more accurate by mixing all three board-sizes than be separating them. Have you read that analysis? It’s been mentioned several times in this thread already, but no one has linked to it yet. I’ll try to find a link. I strongly suggest you read the analysis.

If you’ve read the analysis, and you want to discuss it or explain if something about it seems wrong to you, I’d be happy to read your arguments. But right now, what I see is on the one side, you who repeatedly claims “OGS ranks are broken, you can fix them by separating board sizes” without any arguments, and on the other side, a well-developed analysis showing that OGS ranks are in fact not broken, and are better without separating the board sizes.

In my opinion, OGS is doing a fantastic job with its ranking system. And they’re not sleeping on their laurels. Every now and then, OGS users are asked to take part in a survey where they are asked for all their ranks - their OGS rank as well as their ranks at their national association, as well as their ranks on other go servers. And every year, there is a new analysis posted, which explores how the ranks are holding.

11 Likes

Huh? ROFL.

As I noted above, sandbagging is the overwhelming influence on rank inaccuracy (at least at the kyu level, botting may be more important at the dan level). You are, of course, entitled to ignore this fact, but it renders silly your focus on board sizes. Most people have no idea of the scale of the problem. My notebooks from the time I moderated have about 2,300 names and IDs of violaters, about half of which are sandbaggers with, collectively, many thousands of accounts involving tens of thousands of games.

Thank you for that information. I had heard of the Turing tournament, but had assumed it was a bot competition with bot admins playing in their bot accounts. I had no idea they were playing in their human accounts. That is completely illegal under site rules prohibiting outside assistance. As you say, it is ridiculous and should be ended immediately. Prohibiting X over here and allowing X over there is morally confusing and greatly undermines the rule of law, as abundantly demonstrated by many issues in the real world.

You are justified in your pride. But it is a logical mistake to equate the particular with the general.

They don’t (by and large), and it is.

If a server wants to be good, it should taste like chocolate. And servers don’t have ranks.

Pretty easy to understand: (1) Players want some measure of their skill compared to other players of 9x9 and 13x13, and (2) it is needed for match-making.

3 Likes

:point_up_2:t2:


Other than that, I’m against halfbot tournaments, but I’ll wait and see. If it must stay (I can’t imagine why but who knows), we’ll probably get an explanation, if it must go, it probably will soon. :woman_shrugging:t2:

2 Likes

One reason why it would stay is “because it’s fun”. I’m currently taking part in a rengo with 160 players per team. Are you against that too? Can you not imagine why “it would stay”?

We’ve been playing crazy go variants in real life since forever. It used to be much harder to do it online, because online we can only do what the computer has been programmed to allow us to do; we don’t have as much freedom as in real life. But OGS is making a lot of things possible. Those things are fun. Is there a point in forbidding those fun things?

4 Likes

I will take this as a personal attack, sir.

glove thrown

glove quietly withdrawn

Anyway no, I happen to love rengo. It’s evident everywhere.

(Ranked games with AI assistance are a legitimate concern and nothing to do with fun, I think we’re talking about different things.)

3 Likes

Can we not just make the games of that tournament unranked? It sounded as if both (edit: someone who doesn’t want to be bothered) and you were saying the whole tournament should be forbidden. I might have misunderstood.

3 Likes

Tournaments are ranked games by default, except a specifically sanctioned 25x25 that run at one point.
If I’m wrong about this I’m happy to be corrected.

I don’t want ranked games with AI assistance, unranked fun is fine by me.

3 Likes

I found this one thread with the analysis. I can’t recall if I had seen other threads.

1 Like

not even separating board sizes, they should completely remove the option to play ranked with 9x9 and 13x13, separating board sizes would work for me since it would solve my current problem but since it’s not doable and possible they should just completely remove the ability to play ranked on any other board sizes other than 19x19, and we are not only talking about board sizes, there are simply too many loopholes in the current ranking system, like I said I am making a valid point and it’s not people’s fault that they are exploiting their ranks, it’s the server’s fault for allowing such a thing happening in the first place. a few days ago there is a guy who reach 9dan by playing with 50 seconds total time and winning every game by timeout, it’s not the guy’s fault, it’s the server’s fault for allowing him to do such a thing. there are simply too many loopholes but apparently they are all overlooked

1 Like