Glad we understand eachother.
@Sarah_Lisa
Yes 
@GreenAsJade
Well… I seem to remember that some players had ranks as high as 15dan. are you suggesting these players would beat 9dans by 6 stones 
. Ok… I know you are not. AlphaGo might be able to though, I give you that.
Yeah, now we’re talking more about calibration then the overall principle
So, my thoughts on the topic are…
Given the use these days of elo / glicko rating points, I can’t really find a reason to soft cap “skill levels” when the points are free to rise or fall away from these.
And if we want to take it back to the original definition of one rank being equivalent to one stone in strength, again, I can’t find any logical reason why this system should be confined to 34 ranks… if there are 9 dan players that can beat other 9 dan players even when giving them a stone, why should they not be 10 dan? likewise if there are 25 kyu players that can give a stone to other 25 kyu players, why should the weaker ones not be represented as 26 kyu?
Regardless of whether we speek in modern glicko / elo terms or tradition stone terms… a soft cap at either end simply has no logical justification.
Furthermore, since the bottom end seems to be the point of focus… while all systems everywhere share the 9 dan soft cap, almost every system (while having a bottom soft cap) has different lower soft caps ranging from (jn my experience) 18 kyu down to 30 kyu (and I’m sure other ones exist beyond my realm of discovery). This adds even more doubt as to whether any lower soft cap at all is actually required (IMO).
Oh, why arbitrary? The ‘up to 9d’ scale is not limited to Go. Across all kinds of disciplines 9d meant the pinnacle of human abilities. 14d doesn’t make any sense in that context. It’s as plausible as using a fork to eat sushi. No one can convince you not to do it if you really want to, but … there is simply something very wrong about it. I’d assume that there is also an utter bottom of human abilities, which would stand for the other reference point.
So if we can agree that some sort of an arbitrary limits can be defined, it shouldn’t be that hard to squeeze glicko points in between there. There is nothing wrong with a number 25. It’s as good as 30 or 39 or any other number. It’s just the lower boundary being overrepresented by whole range of different skill levels that causes the feeling of unreliability of ranking system. It should be fairly easy to squeeze whole glicko range properly within any arbitrary scale, so that there is a continuous sense of progression throughout the scale.
I thought that if you have been through an interactive tutorial which not only explains rules but also shows basic capturing techniques (chapel, ladder, net) and fuseki strategy and apply it in your game you should already be above 25 kyu, even if you didn’t play a single game vs human yet.
Am I mistaken?
I had no idea that the idea of nine-dan was used outside of go across all kinds of disciplines. Do you have some examples? (I hope I don’t sound cynical, I’m genuinely curious.)
I guess they CAN be defined, but why should they, especially when it comes to go? The consensus seems to be that there is a lot of meaningful distinction to be found beyond the site’s current rank caps. Why not keep with the current ELO-to-rank conversion and let it naturally determine what the lowest or strongest ranks of players on this site are, instead of pre-defining them as 25k and 9d and then retrofitting the ELO into those boundaries? As has been pointed out, this would change the current dynamic of “one stone per rank”, for the sole purpose of keeping these (arbitrary or not) rank caps.
But wouldn’t it be even easier to keep the current method of ELO-to-rank conversion, which provides the same continuous sense of progression, and remove the artificial caps?
If I understand correctly, your argument is that human abilities have a lower and upper bound, and that the ranking system should reflect that. You feel that 9d should remain the upper bound for traditional reasons, but you agree that the lower cap of 25k is arbitrary.
I’d argue that the reason that the ELO system shouldn’t be fitted to the ranking system is that the best players (which include bots like Leela) are continually getting stronger in ELO, meaning that everybody’s ranks would have to be continually adjusted to make sure that only the strongest players have a 9d rank attached to their ELO. The ELO rating of these players would continue to rise but their rank would stay 9d.
Not only that, but the ranks of players with stable ELO would constantly go down in the face of 9d players continually getting stronger. I think that it would feel much more natural to have strong players’ ranks go up as their ELO rises. The current system of translating ELO into a rank makes much more sense than translating a rank into ELO, in my opinion.
When I started playing on this site, I went through its tutorial, which only explained the basic rules (stones without liberties are captured, groups need to have two eyes to be alive). I was significantly weaker than 25k (like I said, it took me a month to get to 25k). I doubt that I would have even understood a tutorial explaining fuseki at that point. Like most beginners, I spent most of my time playing 9 x 9 go for the very reason that you don’t have to understand fuseki in order to play it. I grant that I might be a slow learner, but there are a LOT of players weaker than 25k (meaning 1000 ELO) on OGS, so even then, I’m one of many slow learners.
When I first read this, I thought that it was a troublesome definition. But actually, I can see that it makes sense, and is perfectly implementable.
However, what this would translate to in our glicko system is that the glicko/ELO would have to have a cap at the equivalent of 9d.
This is the case so that when someone betters the current “holder of the pinacle”, they get to be 9d and the other person goes down. This new “pinacle” raises the bar for everyone.
Then that new person is at the pinacle, the other person is demoted due to their loss, and that person beats other 8ds, who then go down, and the whole “raising of the bar” moves downwards.
But if (as I understand it) you allow everyone at 9d to accumulate more and more ELO, then it no longer represents the pinacle, it just represents a growing pool of people above a certain skill level.
So I think there are two choices that make sense:
Declare that 9d is the pinacle, and stop the ELO growing from there.
Allow the ELO to grow past 9d, and reflect that as a higher dan, recognising that there is no limit in ELO to human potential: no absolute rank set as defining the pinacle.
GaJ
Well, to be honest I did not consider the issue to be a truth or false question, but merely a matter of different observations. That’s why, after reading all the comments, I wanted to share that my experience is different. It is surely highly subjactive to individual players. I would not dare say this is a case for everyone.
As can making it out of 10k to 9k for many people. I do not view this as a reason to add more ranks, that’s how it is supposed to be. I still think that above 20k the skill differences are partially very cosmetic and wins can be very random to a certain degree and just do not think that adding more ranks would help. It just might create (again, just my opinion, no hard evidence) many players wildly oscilating between 40 and 35 for example. Just so we can pat ourselves on our backs saying wow you made it out of 40 so quickly? This to me seems unnecessary. Again, I do not mind it, but I see no reason to “reinvent” go so to say.
AFAIK (and this is without any substantial research) in Asian countries the lowest rank at tournaments/servers is mostly 30k. (at least I have never heard of any official rank below 30). When Europe adopted the system they usually settled for 25 or 20k as the lowest. Do not ask me why, I do not know. Maybe someone more educated might explain… 
Most notably this ranking system is used in almost all Asain martial arts. But is/was used for many traditional disciplines like caligraphy for example. While the starting rank is different for every art/style, the 9d is always pinacle.
Yes, in a perfect world I think so too, but the ranks are kind of “fluid” depending on the pool of users I am afraid… which may cause the uncertanity.
In summary, I donť mind if 40k is created, whatever. But to me it seems totally unnecessary and might just be that people get too attached to their numbers (which should be discouraged in my opinion). I am all for adjusting the Glicko range to fit better between normal ranks if it is way off, but the (approximate) comparability with other servers should also be held in mind. I am afraid anyone who would say I am a 14dan would be lauged at everywhere else 
Thank you, I understand what you meant better now. Although I think my suggestion to reflect ELO ratings with accurate corresponding ranks is hardly a reinvention of go, just a minor change (only visual) to OGS.
Very interesting. Thanks for the education.
I doubt that anyone currently playing on OGS (including Leela) is a 14d (although I don’t know this for sure). The higher the ELO rating, the longer it takes to reach the next rank. In any case, someone who reaches 10d or better definitely earned their rank over here and, given the good reputation of OGS’ ranks (which are stronger on average than the same ranks on other internet servers), I doubt they would be laughed off elsewhere, even if there are some 9d elsewhere that are as strong or stronger.
Some reference:
So - here’s the thing. This is where Go is now different. In Martial Arts, it is fair to say that the pinacle of performance doesn’t change much over the years, and so it makes sense that there is a “top of the scale”.
As it was with Go and Chess.
However, we now know (this is why I mentioned Alphago, sorry!) that the pinacle of Go is far above our current 9d max.
That is why it makes sense to me that Go will break with tradition - a tradition steeped in the assumption that the current masters are as good as it gets. Badly wong assumption.
I do agree that there would likely be some derision. There was also derision of the idea that a bot could be a pro… who’s laughing now?
GaJ
I think 9d shall stay at the top, more is not needed.
Normally the system handles itself - as 9d is not a fixed value, but a value within a player pool.
If a weak 9d looses against strong 9d, he gets demoted to 8d.
If it happens often enough, the ranking system will reflect the new values.
edit:
as for the 25kyu issue - i think that’s fine too.
When players learn to play it soon doesn’t matter if he/she is 25 or 30kyu.
If a player leaves it doesn’t matter
And you find games easier - more opponents to chose from the same rank
There is a huge difference in the ranking system of OGS and kyu and dan grades in martial arts. The latter are marking your advance in practice, are not necessarily based on competition results, but are more like grades you pass in front of a jury. And once you get a grade, you are never retrograded, unless maybe as a disciplinary sanction, and I’m not even sure of that. I got my shodan (1 dan) in aïkido years ago, it took me about 20 years of practice to get there, and even if my skills are not what they used to be, I’m 1 dan for the rest of my life (unless I decide to work my way up to 2 dan).
So, a grade in martial arts is more a reward for the work you’ve done to get there than a measure of your level, skills, efficiency, whatever. That said, my view on this is biased by the fact that aïkido has no competition, no fight with a winner and a loser. Maybe it’s a bit different in judo, karate, taekwondo. But you have also kyu and dan in chado (tea ceremony) and ikebana (flower arrangement) without fights, as far as I know.
So, I must say I was puzzled when I first joined a Go server to see your rank could go up and down depending on the result of one or a handful of games. For me, naming this “kyu” and “dan” is just weird.
Saying “9d is the top” doesn’t even make sense speaking only of human go. A top amateur (7 dan is the max amateur rating) is far weaker than a top pro (9d/9p professional). If OGS ratings roughly track amateur ratings, there’s a whole 'nother world of strength past where the 9d high water mark would be.
Even a 9p isn’t viewed as the pinnacle of possible human ability. Pros, as a group, are getting stronger, and modern top pros are viewed as MUCH stronger than their historical peers. Cho Chikun, one of the top players of his era, and likely the strongest go player in the early '80s, is rated as roughly 300 ELO points lower than Ke Jie and Park Junghwan, who are probably the strongest modern players (and probably the strongest human players ever). Go knowledge will likely continue to develop, and with it, the pinnacle of human strength will increase.
I used to play on IGS somewhat regularly, and I vaguely remember that there were some 10d/11d players from time to time. I think they were generally pros that wanted to play some internet go. I think that, in absence of professional ranks on OGS (though adding them would be one possible solution), it would make sense to add ranks above 9d. If there’s a large discrepancy between 9d players, expanding the ratings to show that discrepancy seems like a natural solution, even if it goes against tradition.
What about having some way of having multiple ranks above 9 dan and having them named 9d, 9d+,9d++,… or something similar.
I’m not sure I like that idea myself but it seems like some sort of compromise between the traditional cap and describing higher levels of ability.
Also FWIW - I think it would be a good idea to have 30k as the minimum rank paired with giving new members the option to state their own rank upon joining so that only complete beginners get 30k.
I think I agree with most of your points and I don’t have a strong opinion of what is the best solution. There are some interesting challenges with scaling the ranks properly that should be approached in a complementary fashion.
I suppose the primary factor should be that 1 rank difference should correspond to single handicap stone statistically. That generally should trump any other considerations. It would be nice if there was some anchoring mechanism that would prevent human players from going above 9d (I don’t mind bots). In my mind any ranking system that is able to fulfill both of these purposes is a good ranking system.
That’s pretty much the current system. It displays strong players as 9d even if they’re one or several stones stronger than other 9d. You appreciate that because of tradition, it bothers me because it feels arbitrary.
OGS actually does have professional ranks for certified real-life pros, for instance: Haylee
In my opinion, professional ranks confuse the picture even more, though, since they scale differently than amateur dan ranks. I don’t think the scale goes 6d, 7d, 1p, 2p, etc in a linear fashion. I think I’ve heard it said that there is only about one stone difference between a 1p and a 9p. And can be it unequivocally stated that 1p > 7d? In any case, I think we should keep our ranking system based on the current ELO-to-rank conversion even for strong players.
I may have expressed my thoughts too briefly. I don’t see a lot of disagreement between what you are saying and the point I’m trying to deliver. The primary aspect is to have 1-level - 1-handi-stone equivalence, which is clearly not the case with current system. ON TOP of that it would be nice to have an anchoring mechanism.