Recognising and fully implementing ranks beyond 25 kyu

This thread is intended to be a partial continuation of the excellent thread started by @Maharani : Please (re)instate ranks beyond 25k and 9d which is now closed.

I would like to limit this thread to discussing “recognising and fully implementing ranks beyond 25 kyu”. The other half of the original conversation (9d+) seems to involve different issues.

I’m not suggesting any change to the Glicko2 ranking system. It already goes beyond 25 kyu. We just don’t display it and possibly don’t use it for match-making.

My experiences as an amateur IRL Go coach mirror those expressed by others in the original thread. ie. That there are significant and meaningful differences in the 25k and beyond ranks.

I believe it is possible for a 30 kyu player to remain at that level and enjoy the game at that level and they should be afforded the same opportunities to arrange games (or tournaments) at their level.

Semi-arbitrarily, I’d extend the system to recognising ranks to 35 kyu.

I would be interested to see a current rank distribution graph of “Active non-provisional ranks on OGS” if @S_Alexander or someone else can provide it, to further inform the discussion.

12 Likes

I guess the first question is: can the system even display and handle ranks beyond 25k?

5 Likes

Only Anoek can answer that for sure, but I can’t see any reason why not: it already tracks Glicko rank down below that, as you can see on the profile graphs.

I have certainly experienced from time to time TPKs having frustration about when is their rank going to change recognising their recent wins, as they claw their way up towards 25k’s worth of Glicko points, so I think this definitely has merit.

As far as I know the arguments against are purely academic … “is there a meaningful difference in skill?”

I would say “who cares, if it achieves the goal of offering people recognition of success in the early stages”.

(Note that my opinion has definitely changed on this topic since it came up last time. It’s been changed by seeing first hand the frustration people feel when glicko drives their rank way below 25k and they have to climb back up with no visible reward)

10 Likes

Why wouldn’t it. As Kosh said, the Glicko2 numbers already go beyond 25k and the app displays the equivalent kyu.

In the UK, the (youth) system starts at 35k and for clubs with enough members it seems you can get a surprisingly robust ranking scale to 40k.

My final thought is that since (as I understand it) all the grades have been pushed down somewhat over time and especially since AI, having the lowest grade at 20 or 25k is probably out of date now.

7 Likes

Because it’s the only reason that this would be stopped? It seems like it’s worth verifying to me…

3 Likes

As far as I know the main reason why ranks weaker than 25k have been excluded thus far has been in the name of match making; the idea that a 30k player would rather have a quick pairing with a 25k player than perhaps needing to wait a long time to find an even game…

I believe accurate data representing the current size of the active population of players with an elo below 1000 would be critical in making the decision to open up additional ranks lower down.

I know the ranks could at least go down to 30k… @anoek would need to verify whether 35k is possible

7 Likes

I am sure that some 30ks would appreciate this in tournaments, especially if I can look at tournaments with a rank restriction of 25k that after almost a week and a half, 3 out of the 6 players are not 25k anymore. They are 17k, 19k, and 22k. This shows that there was probably a large gap in the rankings in the tournament that allowed this to happen(25k vs 30-40k). Someone that is a 40k making it to a 17k within two weeks is almost impossible, unless they are falsely ranked, although 25k to 17k is much more manageable then from 40k(15 ranks higher). I once moved up almost 7 ranks in a day(then moved slowly down for a little bit and ended where I had started a few weeks before). I am not saying it can’t be done, but I am saying chances are that this person was a 25k, even if some others in the tournament are 40ks. I know that some people would prefer to be able to limit the tournament like this because they like to play people that they have a chance against.

Well this was written without looking at. BHydden’s edit. Sorry, BHydden.

2 Likes

I think 50k - 26k should exist, as should exist 10d - 20d
But, 50k should always be able to ranked challenge 25k or they will be unable to ever rank up, its hard to find 41k
same for 20d, they most likely are bots, but what if there is bug and they need to rank down?

3 Likes

I was 25k for a long time, but it is evident I was more accurately 30k. I’d think now I’m closer to 25k.
In my humble opinion as a person in the group affected by the change proposed, it would make a huge difference to comprehension (and morale), please implement.

11 Likes

That would just be extremely confusing in my opinion and would serve little purpouse + think about the HC system, which is now confusing enough already :smiley:


Not that I am against the idea of some reasonable extension (although re-reading it it totally sounds like I am :smiley: ) - I know at least 30k range is regularily used somewhere, but to offer some potential counterarguments:

Though purely academic perhaps - IS there a meaningful difference? I am sure there might be a 27k who can “realiably” beat a 29k in a small club or group just because of some personal differences or strengths/weaknesses between the two of them. But on a wider scale among hundreds of players? Based on my experience and feeling (nothing more) I kind of doubt it…

Let’s say for the sake of argument that difference is not discernable, would it matter? The potential problems I can see with that are: These low ranks will just be volatile as hell. On a good day a 25k will rise by beating a 20k three times and then inevitably fall again when a 30k beats them. Handicaps will be an even bigger mess than they are now and lead to even higher volatility and confusion and it will be potentially harder to arrange matches.

I tried skimming the original thread, but am not sure. What are the arguments for? That beginners will feel better? I am honestly not sure they will. As the system is set up now, growing a few ranks is really an achievement, beacause the skill gaps are there. If you are a grow from 30k to 25k in a day by sheer luck, you will feel good. But when you fall back to 30 the next day, you will feel bad, or potentially even worse. By making the range so wide that it will remove the actual skill gaps, you will just make the “ranking up” something that is no longer an actual achievement, just “temporary luck”.

And a really new player will ALWAYS get demolished regardless of the starting rank.

2 Likes

Hm, it’s not that simple. When steadily I can see that I’m rising from 30k, it makes a difference. It doesn’t mean anything in a day-by-day basis, but it does in a month’s time. And I don’t think it would be that volatile if 27-30k played each other and not “false” 25k who are more probably 24k and almost 5 full k above them.

Especially for a beginner, it is probably much more confusing (and disheartening) to lose from opponents who seem to have the same rank, when in truth they have achieved that rank as true, while their own true rank is not reflected.

Speaking for myself only, I can tell things I’ve improved and make my rank more accurate, so it would really help if I knew they reflected on my rank. For example (and comparing myself and my close rank opponents with my TPK eyes, not some good player’s eyes), with everyone in the huge group of 25k, we can’t tell that learning to recognize a ladder (random example don’t @ me) will earn us solid k’s. We were probably 29k before that and be “true” 25k after that, but it shows nowhere.

8 Likes

Original argument was that game between 30k vs 25k shouldnt usually be played with 5 stone handicap, since games on that level are usually won by “pure luck” and handicaps are kinda unnecessary and confusing. 25k’s often feel very uncomfortable if they have to give handicaps for the opponent, since they are not usually familiar with playing against handis. And if we are not using ranks as basis for handicaps, why would we even show those ranks to users?

I personally dont really care if we bring 30k ranks back (nor if we had double-digit dans) since both extremes are out of my handicap range. I think the current system does prevent the “30k slum” happening again, regardles if we show that rank or not.

2 Likes

I agree with @Gia that TPK’s glicko rank does change steadily over time in the 30-25k range, and that this would be a worthwhile feedback to them about the fact that when they stop losing they are in fact improving.

The problem is that the first learning/orientation matches that TPKs play drive them deep below 25k.

You can see it on their graphs.

Then they get the hang of it, and wonder why it takes them so long to gain a rank.

No other rank-gain has to wait as long.

I think this is the reason why it should be done.

11 Likes

I think if it was the case that OGS ranked games were enforced strictly to be played with proper handicap this would be important. I think though in custom ranked games it’s only komi that’s enforced/auto? A lot of people don’t seem to like handicap games from what I read on the forums.

I’m not sure myself (at my level) which would be a harder game between extra komi vs extra stones to play against. At 25-30 kyu though I imagine it’s tricker to make the extra starting stones on the board inefficient.

3 Likes

The point was that when players who are 25k or below play with eachothers, there is not generally need to have any kind of handicaps, because neither player is actually strong enough to comfortably play as white (ogs is really beginner-friendly server and its nice to keep it that way), and both players usually blunder so much that having few extra stones on the board doesnt really make much difference.

If anything, for players who are just only learning how to play, handicaps might just make it harder to learn about opening strategies. Thus the automatic handicap does not give any handicaps when both players are max 25k.

edit: Note that automatic komi does not work as reverse komi (i hope it would, i’ve tried pushing automatic reverse komi as handicap option for ages!!).
Automatic komi just makes the komi what it should be according the used ruleset, boardsize and possible hadicap (0.5 / 5.5 / 6.5 / 7.5 / 8)

3 Likes

I think I’m confused.

The original argument was that if players are at 25kyu and there’s a reasonable skill difference between them, we shouldn’t differentiate between them in rank because that would have connotations about whether one player should be able to beat another player with x amount of stones. Is that right?

I don’t think x kyu really translates to stronger player can beat weaker player with x stones anyway in the ogs ranks. This would probably be true if all ranked ogs games were forced to be properly handicapped or use reverse komi in some way. From what I understand they aren’t, and there’s a number of users that don’t like playing handicap games anyway.

So you could probably argue that a strong player at any rank might find it tough against a weaker player with handicaps if neither are used to handicap game; or in fact that it could make no difference. I don’t know how to argue either side, and I’m not sure how to interpret what you’re saying for the 25kyus.

Are you saying handicaps are unnecessary for 25+ kyu as they would affect the outcome as much as flipping a coin would (something random/luck based/unrelated), and they just serve to make one or both players uncomfortable?

3 Likes

As someone who has only given others handicaps in 9x9 games, it can be really difficult. I think I won one game in all that I gave someone 2 stones and I got 0.5 komi, and they were probably a 30k instead of a 25k(for reference I am a 16/17k who does mainly 9x9, although I was 20k range for some of these handicap games). If I play someone closer to me and I get 3.5 komi, I usually win(they are around 2 ranks above me usually for this). If a 25k had to give a handicap like that on a 9x9 to a 30k, I think that the 30k would have a better chance of winning, even if they only sort of knew what they were doing. I think that sometimes in 9x9 the handicaps can make it difficult for the people that have to give stones, and at least in my opinion, the 3.5 points don’t matter(same outcome with 5.5 points). I also don’t really understand handicaps in 9x9, so forgive me if I am wrong, but I think that a 25k shouldn’t have to give 2 pieces and have 0.5 komi against a 30k. The 25k is unprepared to give two pieces, and the 30k, if they know how to win(you need territory and captures to get more points than your opponent) could beat them because of the handicap.

1 Like

Yeah, i’m saying that games between 25+ kyus are basically like coinflips, with or without handicaps xD

But like, the reason why we have numerical ranks in the first place is to find similarly ranked players. When ranks differ, we go players usually compensate the rank difference by giving the lower ranked players handicaps. Even if non-handicap games are more common on ogs, thats still the normal way of setting up matches all across the go world.

3 Likes

Imagine that there were only 10 ranks: kyu, and then 1d through 9d. The kyu players see themselves never gaining rank, and want to implement some gradation so that 9k-1k exists. Some objections might be that sdk ranks would be highly volatile (which is true, compared to dans), or that sdks don’t know how to play with handicap stones (which is also true, compared to dans), or that sdk games have so many mistakes in them that the games essentially end up being a coin toss anyway (which is also true, compared to dans).

If the glicko can handle it without giving weird results best expressed with complex numbers, I can’t think of any more compelling reason not to show those ranks than I can think of not to show kyu ranks.

12 Likes

I think you’re really stretching there.

On 1k-9k range it’s not true. One rank difference gives 70-75% win change to stronger side. More is like 80%+. You tried to sneak “compared to dans” there but compared to dans or not this is not how coin tosses work.


With new ranks it’ll be harder for beginners to find games, right. Since 9 rank hard restriction and 3 rank default in automatch.


Anyway, I think this is pointless. Is this really such a big problem anyway?

5 Likes