Perhaps the AGA could allow players with no established AGA rank to register with their OGS rank (or another server), and apply a sensible conversion.
2 Likes
There’s only two site admins/developers really. There’s a couple more people that contribute largely to the code. None of those groups include me.
If you want to quote which part you have as a problem I’m happy to reply with my thought. They’re just my thoughts though.
The logic presented wasn’t very convincing in my opinion though. I’d love to be four dan, but we shouldn’t shift the ranking system from 3kyu onward to make me a 4d, even if I could achieve 4dan on fox.
In the past in Ireland, if you’ve regularly played someone but their tournament rating is outdated by at least two ranks, but close to you, you just declare them as that rating. You don’t necessarily take some random online rating at face value.
You certainly don’t opt to change a large chunk of the OGS rating system, rather than applying some conversion as
@martin3141 says here.
We had a player declare that they were like 2d or so in China one time at an in person tournament, so I played a quick game with them with them giving me handicap. I won easily. Then we played an even game and I won easily. I inferred from the kinds of mistakes they made they were more like 10kyu and they had a much more even game around that level.
You don’t always take peoples declared ranks at face value. If someone is massively overperforming for a declared rank, you can also, and probably should, adjust their rating before submitting it into a rating pool like EGF, so other people’s ratings aren’t affected.
Someone else can point at the chess comparison if they want. It doesn’t make sense to me. The pool of chess players is much larger, so the rating ceiling typically is going to be larger. We can globally shift the ratings to make 2000 be the bottom 10% if we want, and keep everybody the same relative to each other.
7 Likes
I’m 9 dan on OGS and 4 dan in EGF, so OGS is obviously massively over inflated and everyone’s OGS ranks should go down by 5! 
7 Likes
Ok so first off my bad - seeing the moderator badge, and the style of your communication, I thought you were speaking for OGS. Thanks for clearing that up.
I think OP damaged his cause by adding that little joke (1: It will make me happier lol), also the rest of point 1 is not quite clear.
I also think the thread got a bit side-railed by the example given of bringing the OGS rank to register for tournaments, which also got a few replies.
Ignoring the above, I realise now the attempt to anchor the OGS rank to EGF and AGA and the figures quoted seem to kinda support that.
Still, it would have been nice if the conclusion had been a bit more spelled out rather than not replying at all to the message by square_fuseki, one of the few to bring facts and figures to what seemed a lot like an opinionated debate.
And as for the main point, shifting everyone up by two ranks may not mean much as everything is relative, but by taking that logic to the extreme we might simply decide that this server has ranks from 60 kyu to 30 kyu or 23 kyu. But I believe that there is a bit of value in granting a 1 dan rank a bit more generously than currently done. Not only is there for many players an emotional connection to the idea of achieving “1 dan”, but in terms of “skill on the board”, OP did address this quite practically by:
What I consider Dan level is when every move makes a positive value to the game. Maybe not the most efficient but positive.
If we keep being content with OGS ranks being “hard” as they are, we do end up having to prefix our rank with which platform or association we got it from, which is tiresome. As opposed to the small difference between Japanese and Chinese counting rules, where one is aware that there may be a 1-point difference, but when you state a score, you don’t normally check which ruleset was used (unless the game was quite close).
The intention was that OGS 1d should be intermediate between EGF 1d and AGA 1d, so for dan players, the rank difference between OGS and EGF or AGA should be less than 1. The OP is AGA 3d and 1-2d OGS, so the difference is 1.5 in his case, but one example isn’t enough to make statistics.
3 Likes
That study is really not convincing. It has very low value considering the pool (like 100 players).
It’s hard to not be in a opinionated debate as you said.
You should consider what is rated at first. federations like AGA and EGF use the frame of tournaments with categories and such.
OGS use any games with the will to be rated, be a hyper blitz, a 6 months, a 9x9 …
I’m not saying the idea is wrong but let’s understand that there could be some differences of appreciation (value) between different objects
6 Likes
That’s no problem. It is confusing, I could probably add a tagline or something that clarifies things.
I generally would also hope to make OGS better so I’m happy to contribute to discussions on it, but I’m also aware from the small contributions I have made, that big changes need to be well reasoned out for them to be made. Generally the developers don’t want to make a change and then have to fix it retroactively like a month or 6 months later.
Sure, initially I was replying to @Jon_Ko who relied to me beforehand, and I haven’t had a chance to look at the survey again recently to see how significant the mappings are to have any sensible comments on it.
I think the main issue with it is that they’ve very small sample sizes when we narrow it down to the ranges being discussed in the OP.
In theory OGS has built into it a setting to put in your rank on many other severs.
If we incentivised people to fill that in, and maybe threw away outliers (my WBaduk rank will be way out because I haven’t played there in a long time), then we could do quite a large scale collection and comparison.
I understand the point. You could go the opposite way and have a system like the Japanese system, you can pass a test or have an organisation like the Clossius Go school award you it.
The thing is each rating system should strive to be self consistent, and we shouldn’t just make a particular rank easier to achieve because people have some emotional attachment to it.
OGS in particular has tried in the past to tie ranks to winrates with handicaps statistically, and if mess around with awarding 1d based on merit or something it’s going to mess up that consistency.
You could do something like give players a badge to show they’ve obtained a rank, regardless of whether they’ve held onto it long term or not. I don’t see any issue with that.
Unfortunately, regardless of what OGS does or doesn’t do, we always have to do that, because Fox is both the most popular and the most inflated of the Go ranks at high level. People, especially if you’re offering lessons will be much happier to quote a high looking fox rank.
2 Likes
I think @jlt, @Jon_Ko and others have suggested there could be some deflation in the ranks over time.
The ratings repo, the proper one using python is available for people to test at
That would be one way to test out an effect, like if having some percentage of sandbagging players in the system was leading to noticeable deflation, or playing too many handicap games etc.
5 Likes
Some people pointed out to me today on stream that some of my opponents are also playing better than their level because it is me. This can be because of botting, logging in to an old account to play me, or trying much harder to win since it is a streamer. Also streaming adds to the difficulty making it harder to play your best.
So it might literally be because it is me that the kyu players I’m playing are stronger than normal lol.
10 Likes
on the other end, there are those who aren’t streaming who have felt some sort of rank deflation. For whatever reason, be that “the level of play is getting stronger overall” or “there are sandbaggers / unwilling to reach dan / cheaters”, this seems to happen.
I do wonder if those who are, idk, 3k and stronger, play on faster time settings overall compared to sdk, compared to ddk. Would there be some effect there? Perhaps also, the change to the preference for 5m + 30s (and the incremental version) from 20m + 30s has changed the landscape.
For me, the time change for matchmaking was an adjustment for me. I used the time to read-read and center myself. I have opted to play 10m + 30s on custom these days. Old irl tournament play gave something like 45m + 60s. Look at what go teachers advocate and what online leagues set for their players, they are supposed to play on higher time settings.
I understand that we “don’t always have time for that” and that other servers play on 5m + 30s. I do think this is interesting, coming back as the prodigal go player (due to the 100 games challenge from Clossius this year lol).
6 Likes
From skimming the discussion the main problem seems to be there is no standardized rating system or agreement of what a “dan” should be. This doesn’t seem like a problem OGS can solve alone.
But something they could try is to align their ratings with those of an other server. Reducing the number of competing standards
A good tool converting between the ratings on different servers would also be valuable.
5 Likes
A nice tool to compare different server ranks is gome.at.
It also gives an average rank. Sadly there is no converting in between ranks, but I guess that’s also very individual.
Here my example:
2 Likes
I’m find it hard to believe you seriously suggested adjusting the entire OGS ranking system based not just on your personal experience, but your experience playing whilst streaming and chatting with viewers.
5 Likes
Love that page! It is also precisely the type of news/events aggregator that I was looking for recently. Thanks for sharing!
3 Likes
First of all, thank you for posting. I saw you broadcasting yourself playing. I asked the same question. I still don’t understand if I’m Dan 1 OGS or Q 1 or Dan 2 in OGS. How much am I worth in Europe? Before the current ranking in 2014-2019, Dan 1-2 was considered Dan. Dan 3-4 is today. And again, it confuses me. So please, something from programming, give us an explanation of what was in the past and what is today. You are right, there are some players who are also Dan 2-1 K in OGS.
Copied from my response on Discord:
-
I’ll jump in the actual thread but basically all rating systems have different strengths, weaknesses, and even goals that are antithetical to others . So trying to just solve one with the logic of another is a no-go. You can prefer the logic of one to the other but at the end of the day, that’s all it is. I honestly think ratings lost all meaning below 25k. I think the data we do have on such games is limited and bad and hard to draw conclusions from. Likewise, ascribing meaning to thresholds with such precision is a fools errand.
-
[5:14 PM]
My only desires to switch to glicko2 for the AGA are (1) quality of life, the things we can do having an actual interactive and well-understood system and (2) AGA ratings work off some assumptions that cannot be true and therefore produce just as many issues as they resolve, based on the player pool. And since our rated games pool is low, I feel like it does more damage than good, or is at best a push. I cannot imagine an advantage, though.
Jacob Upland (Iowa 1d) — 7/6/2025 5:21 PM
To be specific: the AGA system goal to equate 1 rank to 1 stone to strength is both antithetical to the goals of other similar systems and… kind of an impossible task to quantify evenly across ranks as is. But it must solve for this variable so it will, so the results of ddks 3-stone games, for example, will echo into Dan ranks. Because they’re supposed to be the same. But we all know a 20k vs 23k with 3 stones is not going to have the same win distribution as a 2d vs 5d in any sane world. But the equation must solve for these being equal, with the data presented. There will be lots of noise from that.
To add to this, I’m like 5-6dan in Japan but… I played 1dans at the NHK who bordered on double digit kyu. Because there, rank is just a thing you eventually “get” if you play long enough. It has nothing to do with statistics and little to do with strength, you play long enough, they’ll call you 1dan.
If we are equating 2000 to 1d, 2000 is much better than “when every move makes a positive value to the game” on average, yet they are also still occasionally making massive blunders and several minor misreads. So it’s kind of wrong in both directions. They can be as accurate as top players for very brief stretches, but just aren’t as consistent and demons can creep in. I just can’t make this conception work in a way that doesn’t also work for sdk players. I feel like they’re always doing something with their moves, they are no longer just occasionally passing, they just misread, play slow, etc. But its very rare they play very good sequences. A 2000 on lichess will book perfect walk you through a mistake many times.
3 Likes
I assume what the AGA is doing, is that it has some numerical rating coming out of a model, and then it scales in a way that it can assign integer labels that we interpret as a kyu dan system.
All in all though, ranks are just fancy labels of ratings, unless they connect to some aspect of the game. You could call someone 2000 or 1d and it makes no difference, the same way you could call them brown belt.
The extra freedom you tend to have in the kyu-dan system is that when you want to assign an interval of ratings to a specific bucket like 1dan or 2kyu etc, you now have a parameter which is the size of the interval or the bucket.
If you make that interval a constant width, either in the original ratings or after some linear scaling or something like that, then I agree that maybe 20vs23kyu and 2d vs 5d might not have the same win distribution.
However you allow the intervals to change width as a function of ranks, then you can account for this difference. You might have small intervals for tpk or ddk players and wider rank intervals at high dan, to accomodate for the difference in skill needed to win with extra stones for example.
I noted this about OGS previously here
Again I don’t really know what the AGA does, so maybe you’re generally correct about your assertions about its system.
4 Likes
I don’t currently have issue with the math. The issue I have is the perception and noise. I think 1k and maybe even 2k, could be the start of Dan level because I think they might deserve it. I want Dan to be achievable for people as it is a very big milestone for lots of players. Also, it can reduce the perception that no Dan players play on OGS. (This literally could be the reason Fox makes it so easy to get to Dan level.)
For the last several days I have been playing 5k-2D and finding very interesting games. Some 5k/4k give me just as hard of a time as 1D-2D sometimes. I did play a 5 kyu a second time and that went horribly differently lol. I think the type of game matters, which I already know of course, but it is interesting to see in practice. In addition to just the type of game, I am finding there is some inconsistency to ranks. Some 4k are stronger than other 4k. My theory is because there are too many ways to be a 4k. Blitz, live, corr, handicap, no handicap, & size. I’m not sure this is a math problem that can be solved…
But this is getting a little off topic as my point was to lower the Dan entrance. (That and the 25k problem.)
I think (hope) more people would play games if Dan was easier to reach.
3 Likes
This is the exact reason for the “prompting dan ranks” started to appear in CJKT regions in recent decades. And it was the sudden popularity of Go that started this trend, where parents saw their kids wanted to play Go since the early 2000s, and the Go schools saw the opportunities to “promote” their internal rankings and sending the kids to competitions as early as possible for a faster “turn over rate”. And the associations also started to see the benefit of gathering more certificate fees in the name of promoting Go and dropped the threshold for promotion as well as allowing self-report ranks all the way up to 1k level. (hence in a 1k competitions we saw players from strong ready to be dan players from the previous generation, to nearly novice in the same bracket).
As to how effectively this strategy really was, I’d say it was pretty successful in the beginning, and the Go population surged. However, we likely exhausted all the potential students in a very short amount of time, and then started to suffer the consequence, where the inflated dan ranks matters less and less, and we end up with paying to be dan, than playing to be dan. And in recent years, this trend started to reverse (more rigorous and disallowed self-reporting ranks but the lowest 15k to start). There is a balance need to be reached.
10 Likes
Maybe it’s a big milestone because it’s difficult to achieve? Then making it easier also makes it a less impressive feat.
12 Likes