Glicko-2 does not require accounts to start at 12 kyu. So why do we do that?

That is alt sandbagging, which has long been a huge problem for TPKs and DDKs, and to a lesser extent for SDKs.


My suggestion wasn’t to force them to do some tsumego, it would be an optional extra to help you asses your rank if you choose. The workflow would be more like:

  • Know your rank? Select it from this dropdown (with some texty descriptions so people who don’t know what kyu and dan don’t pick nonsense)
  • Don’t know your rank? If you are a beginner pick 30k.
  • Have played a bit but don’t know your rank, do these 5 tsumegos and we will assign you 30k / 20k / 10k / 1k/ 5d based on the first one you get wrong (the first one would be just capture a single stone in atari).

I love the typo from a heavenly dan to a lowly kyu. :stuck_out_tongue:

(I know it’s a typo, I’m my usual cheeky self)


In my opinion nothing will change if we mix all and propose to change things which are not badly working. I’m refering here on the system we have for not beginners players.

So we should focus only on how to welcome beginners better. For that you need to determine who is beginner and i don’t really see better friendly way as to ask directly when someone register.

The welcome process could be more elaborated as now. I too don’t think a test is the way.

The way is

  • Let them play together (shouldn’t this be obvious?)

  • Give them an opportunity to have a kind of coach (we have the means to do it).

  • Let them be automatically registered as 30k at first (why asking them to do it?)

Starting from that we can build a specific system for them to make them experiment the best beginners’ place on internet.

Mods may not agree with my perception so It’s a kind of question about the sandbagging problem. Is there so many beginners that we could not assign a tutor/coach for each of them? I mean a 15k or stronger who on his will take care of the first steps, helping to close boundaries, understand the 2 eyes, the scoring… Same time isn’t this a very good filter against some sandbagging intention?

On a side note i think live games are much more appropriate as correspondance games as it was suggested.


I’m sorry for misunderstanding what you wrote. There was no intention of quoting you out of context. I see now that both statements were in regard to the Glicko-2 algorithm. I took the first statement (the one quoted) as referring to what is happening on OGS (which has caused tremendous confusion in multiple threads over months).


No worries pal :grin::+1:

1 Like

Is this a sign that lots of people erroneously set their rank as 30 kyu (whether on purpose or by accident) as suggested? Or is this actually a sign that we genuinely have a lot of beginners? Without some more detailed analysis of the data, it’s hard to say. :man_shrugging:

These are important issues. Thank you for raising them :slightly_smiling_face: To address some of these issues, I suggest the following:

  • Clear guidance on how to select initial rank (avoids beginners thinking 6d is appropriate)

  • New accounts declare rank (if known) to set initial Glicko rating but rank still shows as [?] until deviation drops below existing threshold (avoids ego chasers, etc.)

  • Option to reject games from [?] accounts (allows existing OGS dans looking for serious games to avoid new accounts erroneously set to Dan and allows established 20-30 kyu accounts some protection from sandbaggers)

This is a really good idea :smiley: Coupled with my point above that even with declared rank it is still high deviation and shows as [?], this could work quite well. It could even be a couple of questions on the registration page. Something like this:

Ranks allow matching of player abilities for fair games. Most new accounts start with a provisional rank that shows as [?]. This should settle on an appropriate value automatically after a few games (5-6 on average). The questions below may help to reduce the number of games necessary to get a confirmed rank.

  1. Do you know your rank? Yes / no

If yes:

  1. Enter your rank here: [rank drop down] [association / server drop down]

  2. Enter link to your association / server profile: … (This will allow moderators to confirm your rank and skip the [?] stage completely)

If no:

  1. Select your level:
  • Beginner
  • Novice
  • Intermediate
  • Advanced
  • Don’t know

(the above would map to, say, 30 kyu, 20 kyu, 10 kyu, 1 Dan and don’t know reverts to current behaviour, i.e. 6 kyu)

This is similar to @Uberdude’s suggestion :+1: but without the need for any tsumego :wink:

This also covers some of the sign-up process parts of @Groin’s ideas. I also like your idea of assigned coaches for beginners but I think that would be difficult to implement. You would need volunteers and someone to manage assigning of coaches to beginners. I like it in principle but I can’t see if working in practice :slightly_frowning_face:

Both. Ogs is very popular place to start playing thanks to different languages and working on a browser. Theres constant flow of new users who have just learned the basic rules, they all stack at the bottom of rank pool and they are all improving their skills with every game they play.

There’s some good suggestions. Ogs indeed had some sort of “how experienced you are” thing when registering but it got scrapped quite fast, i dont know the reason why. Maybe if @anoek has time he can tell his opinions on that.

I feel like the biggest problem is that we’re forcing those established 13-8 kyus to play constantly against beginners. Many people feel its “waste of time” and they rather just resign from the game when realising the (lack of) skill level of their opponent. I assume this doesnt feel very good for the new users either when they would like to play and learn, but their opponents give up just after few moves.

1 Like

Easy. Create automatically a coach request list on the main page and visible only for 15k + players.
Click a name will message a "Greendragon would like to be your personal tutor. Do you accept or would you like another one? "
Then the tutor’s name could be written on the player’s profile. (And the tutored name on the tutor profile).
You then can have the dismiss button there too.


Why does OGS do that? Because OGS ‘development’ means adding features that the developer thinks are “cool” instead of fixing problems (and outright bugs) reported by users.

That’s why I stopped being a (paying) supporter of OGS and moved to KGS.


I can’t disagree more. The last big update gave us a stronger KataGo interface and the ability to let the AI judge variations, both features that have been requested by people. The updates before that revolved around fixing the ranking system, which used to be the most criticised part of OGS (apparently successfully, since most of the criticism has ceased). The tournament system is being renewed and expanded, which is also frequently critiqued / requested.

Most “cool” features (like the Joseki dictionary, the rating graph per game, the colours of the game list, accessibility options, etc.) are done by volunteers from the community, not by anoek.

Furthermore, a lot of development has to go into making sure the server works at all, not in new features.
A lot of the work anoek does is more or less vital for the server, but not as visible from the outside. The number of new players has exploded since the pandemic started, which meant a move to a larger data farm. With AI analysis, the need for servers to do the computations also increased. Last week facebook decided to drop OGS, which required an immediate fix. There’s apparently a finance problem as well that needs urgent fixing with a strict deadline.

Most problems and bugs aren’t fixed because there are way more important problems to be fixed first, not because anoek has only time for “cool features”. If it was all about getting new cool features, we’d surely have rengo by now.


Well I don’t see much of this thread

added, and we’ve had the coolest ideas in there.


Communicating was lacking and i appreciate that although it seems to come a bit late.

Maybe pausing the development of a new tournament system and starting to solve the growing list of bugs would be of better priority?

a troublesome bug is a one of the newest.
I mean losing trust in the system is the worst it would be a shame when organization like EGF start to use OGS.

If OGS is growing like that, associate one more developer?

The criticism are still here coming from beginners users now. But for other users It seems the new system is a success.


The new tournament system is partially there to fix bugs with the tournament system, though :stuck_out_tongue:

In the end, anoek decides what to work on and when, I’m just mentioning the tournament system, because that’s one of the things that was being worked on recently. It’s also a good thing to work on considering more and more clubs and associations move tournaments to OGS.

This isn’t a bug, but a lack of a feature, actually. There’s no notification from games that are in scoring, and they don’t seem to show up by clicking the move counter. I agree that this is an important thing to fix, and anoek is aware of it too, but like I said, currently there are behind-the-scenes things that take priority.


Hear, hear!

…and with all these dumster fires going on, anoek JUST took time out to resolve this other recent bug (a real one this time, not just a tired mod missing the fact that scoring started and ended a whole day apart)


I’ve just realised something else. The volatility bug was that OGS used the value 0.6, whereas the Glicko-2 algorithm says to use 0.06. That’s why OGS ranks used to fluctuate so wildly. This has been fixed now. But look closer (emphasis added):

So actually, if we set the initial volatility slightly higher than 0.06 (but obviously not as high as 0.6), then the initial rank adjustment could be made to work like a bisection, which would get it to the correct rank even faster: 6 games max, as opposed to 6 games on average, but potentially as quick as 1-3 games depending on the rank required. Also, the initial phase would be more a mixture of wins and losses, as opposed to almost all games of one result as I think is the case currently (e.g. “5-6 losses” required for beginners), which would be nicer for new accounts trying to establish rank.

Furthermore, from previous discussions (e.g. the rating update announcement) I got the impression that volatility was a constant on OGS, but again, look closer at Glicko-2:

So it should be a per-player variable that changes along with the rating and deviation. Is this the case on OGS or is it a constant?

1 Like

It is a player variable, but it stabilizes around 0.06.


Oh good :slight_smile:

But still, maybe the initial value could be made slightly higher to help initial ranks to get to the right place more quickly…

Just throwing in here, as an actual 30k beginner earlier in the year.

Given the option to set my own rank I would not have known what was correct; though I’d happily pick a ‘beginner, learning how to play’ option.

The first thing I did was play noob bot and amybot; bear in mind as a new player, I don’t even know how the interface works, so don’t want to risk playing an actual human. I promptly lost, and my rank plummeted appropriately. So definitely if the current starting rank disallows play with low ranked bots, that seems like an issue.

Regarding tsumego to set rank; I could see how this would be a hassle for a dan player, but at 30k this could even be fun (functionally a buzzfeed quiz :rofl:).


This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.