Recognising and fully implementing ranks beyond 25 kyu

[Kaworu_Nagisa]

Has there been any sort of decision or anything regarding this topic by OGS staff?

[flovo]

@Kaworu_Nagisa

Has there been any sort of decision or anything regarding this topic by OGS staff?

@anoek

During our analysis we’ve seen that players whose ranks are less than 25k are very difficult to adequately rank in the sense that handicaps at that level are not a reliable way of accounting for strength differences. As such, we’re now only assigning ranks to players who are strong enough to play with handicaps, which seems to equate to being about 25k and above. The only real effect of this is that when two players whose’ ranks would be less than 25k play each other, they will play without a handicap.

[Gia]

That was 3 years ago, and I’m pretty sure there are enough convincing arguments here that the matter is worth revisiting.

[Gia]

@Kosh

Unfortunately 25k and lower players are even less represented on forums than the average.

Also this is a very important point that should be taken into consideration. Most new players won’t even know how to ask or what to ask for regarding this matter.

[Kaworu_Nagisa]

People complain about it all the time on discord. It’s actually the biggest topic on the OGS channel… “Why can’t I rank up past 25k?”. “why are 25ks so strong?”.

[Kosh]

Thankyou for the link flovo.

@anoek

The only real effect of this is that when two players whose’ ranks would be less than 25k play each other, they will play without a handicap.

• As noted above, it also makes it harder for new OGS users to perceive their progress.
• It makes it harder for any player in that category (below 25k) to view the relative strength of an opponent who is also below 25k.

• It prevents those in that category from using quickmatch prefs or custom game prefs to the same extent as other users. eg. a 26k player can’t specify
  24k-28k for opponents in a custom game.
• Similarly, custom tournaments can’t be arranged where either or both of the rank restrictions would be below 25k.

Disclaimer

[BHydden]

@Kaworu_Nagisa

Has there been any sort of decision or anything regarding this topic by OGS staff?

@Kosh opened a ticket for this issue on GitHub, so if we don’t get a reply here there should at least be on there at some point

[Eugene]

@Kaworu_Nagisa

Has there been any sort of decision or anything regarding this topic by OGS staff?

To be honest, I think the timing of this is such that this topic is not high on the agenda.

At the moment, the top priority is restoring ranks after the disconnect drama - work which will bring other much needed improvements (such as the ability for moderators to annull games more "the most recent ones, hopefully).

I think that when this is out the way, it might be a better time to ask again…

[Gia]

@Eugene

disconnect drama

I will not let the opportunity to call it “The Cor-Corr Incident” go to waste.

For people with no imagination

(corona-correspondence)

[BHydden]

@Gia

The Cor-Corr Incident

So shall it be written, so shall it be done

[DVbS78rkR7NVe]

Long time ago anoek said that he’s going to revisit ranks at some point.

[Kosh]

Before I call, ‘Alea iacta est’, I think there is one more important point to make in favor of the proposal. I just hope I haven’t left it too late in the thread to be noticed…

Having a minimum ranking (ie. 25k) carries with it an implied judgement or standard. Perceiving this minimum, a new user is apt to think that they are expected to quickly surpass this standard or they just, “Don’t Get It”. They may even feel that they are not welcome at OGS!

I hope and believe that it is not the intention of OGS to portray an elitist attitude to new users but that is what we are in danger of doing. I think that is why I am so passionate about this issue.

I believe that Go can be played AND enjoyed across a wide range of skill levels with or without an active desire to improve. It is easy for those with a natural talent for the game to forget what it was like to regularly not recognise atari.

Frankly, I am more concerned with how new players feel at OGS than whether or not the auto-handi works just right, which ofc people will never agree on anyway.

Alea iacta est

Cookie

@Eugene

ask again…

[Eugene]

I happened across a possibly more useful TPK game history here:

https://online-go.com/player/737363/

Here is another person who was driven deep into >25k … down to 740 points (25k is about 1000 points).

They stayed there at 740 points (~34k ) learning stuff for about 20 games, and all these games were against other TPK.

About half of them were amy-bot beginner, 24k, which at first I thought spoiled the sample.

But then the person learned and started winning more than losing - including transforming from losing in a big streak against amy-bot beginner to winning against it.

On May 1 they were right in the balance, playing ~70 9x9 games with 50/50 win rate.

Then they clearly improved again and on May 2 they won many more than they lost - their rank climbing comensurately.

… and the whole time they have been showing up as “25k”. According to rank, they haven’t changed the whole time…

---

(I fixed up the actual numbers based on flovo’s info in the next post)

[flovo]

@Eugene

down to 740 points (25k is about 1900 points)

740 ≈ 34k and
1000 ≈ 25k

[Maharani]

When I created the original thread, I incorrectly assumed that this was simply a cosmetic issue of correctly displaying the rank corresponding to the player’s Glicko2 rating. However, it turns out the reason it’s statically displayed as 25k is because ranked games can only be played against players with no more than a 9-stone rank difference. If the Glicko2 rating was accurately displayed as the corresponding rank, apparently the player pool for below-25k players to play ranked games against becomes far too small.

I joined OGS after learning the very basic rules and immediately got beat down to Glicko2 623, which IIRC translates to about 40k*. After that, it took me 106 games to reach “proper” 25k*, which roughly corresponds to Glicko2 1000. Here’s my rating graph from that period:

My current opinion on the matter: I agree that the best way to tackle this is to display ranks correctly and remove the restriction that non-handicap ranked games can only be played within a 9-stone rank difference. As someone else said, I’m sure Glicko2 is well-equipped to deal with larger rank differences.

*EDIT: This was prior to humble rank, not sure how these Glicko2 numbers would translate now as far as displayed rank.

[flovo]

1. It looks like there are enough players below 25k to get games in a decent time even with ranks reaching as low as 35k. This is assuming, not most of them playing only bots.
2. While glicko2 in theory works for any rank difference, it gets worse when too many games in a row are between players too far away from each other. I encountered some problems there when experimenting with glicko2 for myself.

---

Btw. I’d like to get rid of the kyu/dan ranks for range selection in even games and use ±glicko instead.

[Kaworu_Nagisa]

@flovo

for range selection

My only concern is that no one really knows what their glicko rank is/means. Is there an effective way to communicate the rank difference without simply stating the number?

Also, would it include the error range too? So if I was 1000 (+100, - 100), could I get into someones 1100 restriction?

[Haze_with_a_Z]

Another question, would it use the overall, or the 9 different subcategories of games(size and timing). My 19x19 blitz rating(1560 ±260) is drastically different than my 19x19 overall, live, and correspondence ratings(all in the 800s range). Would that mean that I would be playing a much stronger opponent if I decided to create a blitz 19x19 game than if I decided to create a correspondence 19x19 game?

[Maharani]

The subcategories are only for informational purposes. Only your official rank is used for matchmaking.

[Haze_with_a_Z]

No if we changed out of using the kyu/dan, what would happen in that sense?