I think the 13k default rank is doing harm


The only disadvantage of the solution is the psychological aspect. If we introduced the solution, then overnight everybody would experience a sudden drop. For example my humble self would need to appreciate rank of 2d instead of 3+ where my ego sits today.

I did a quick math and it seems that with the current settings, the humble rank for newcomers with 1500±350 glicko would be ~21k which feels just right :slight_smile:


Some good and bad news.

The good is that ranking conversion is in the open-sourced (UI) part of the code, and anybody can fix it! Yay! https://github.com/online-go/online-go.com/blob/devel/src/lib/rank_utils.ts

The (slightly) bad news is that it won’t be easy as the use of conversion functions is spread around the app (it doesn’t seem daunting though).

EDIT: So (un)surprisingly there are few challenges that go beyond just presentation layer. Some of the functions that run in the backend might require adjustments as well… Some examples I can think of:

  • am I allowed to join this tourney
  • am I allowed to accept the challenge

I removed part of original post to not set hopes too high unnecessarily :slight_smile:


Do you think that the scaling to kyu can be permanently offset by - say- todays average uncertainty, so the transition doesn’t dent egos. It’s easy to joke about egos, but it is a real factor, an we don’t really want OGS ranks out of wack.

My impression is that most regular players (that I encounter) have an uncertainty of 1.5-2, so a 1.5 humble-offset seems reasonable?


I’m a bit skeptical… What about attracting other high Dan players who have to force their way through the ranks?

One of the most notable OGS weaknesses is how hard it is to get a game at the high Dan level, so it stands to reason that we shouldn’t force them to feel like they’re wasting time here, and it doesn’t completely fix the beginner/not accepting games problem.

What percentage of new, low ranked beginners on ogs are absolute beginners? I’ll wager it’s a fair percentage, meaning that some DDKs still might get aggravated by this (and the sandbagging).

I admit, humble rating lessens the problem, but it doesn’t remove it the same way that GaJ’s provisional rating does.


This is a complex topic and while I’ve shared a personal experience in a prior post, I do not have the depth of knowledge necessary to solve it.

That said, I’ll throw this out as a brain tease on which to build. It’s stated someplace (I think) that OGS gets 1500 new players a month…if true, that’s 30 per day on average. Perhaps they can all be matched against one another exclusively so that they can play other new players, or newer players with, say, 3 or fewer games (not necessarily only enrollees from that day). This is like provisional system now, but keeps it inclusive for just a very few games. Until they’ve played other new enrollees, they cannot play others here, ask for challenges, be in tournaments, etc,. In short order (three games or so), the 1 dan would be distinct from the true 13 kyu from the new 25 kyu beginner versus the “just dipping a toe, goodbye now” quitter. After three games against other new players, _then_they could open up to a “provisional rating” that can play more openly on OGS but the rating is a better estimate of their true rating than simply labeling 13 kyu and leaving them to the sharks in open water. lol Win all 3 games, first provisional Glicko rating is 5k; 2 of 3 wins, it’s 10k; 1 0f 3 wins, its 15k; 0 of 3 wins, it’s 20k…for example. Some will ask, “What if three dan level new players happened to draw me and beat me but I’m really not 20 kyu?”, and the answer is that could happen, system not foolproof, but on average, three games would give a better estimate than random 13k assignment. Finally, some might argue that overall, playing three games will simply come to the average of all new players playing, which will be around 13k, but I think this system would provide (a) games for those that simply cannot get games now because no one will play new players, (b) sort out the quitters, who will timeout or quit altogether after a game or two, and C) still, the dan players will tend to win all three games (since there are fewer new dan players signing up for OGS compared to newbs) and the brand spanking new player will tend to lose all three games, so I think it does serve the overall purpose of getting them a new “rating” that is closer to reality than giving everyone the average 13 kyu.

I could be wrong that this would work–I am throwing it out for your building upon or discarding. :smile:


It would as easy as adding another constant to the ranking formula. The downside is that it adds another constant to the formula. :wink: I suppose it could be left as an ‘implementation detail’ until when this idea gets some real traction.


Fixing the UI part was easier than expected :wink:

Obviously this is just a proof of concept and only @anoek knows how much more work might be needed to do it properly. It does seem however that the code is structured well enough to take this kind of change lightly!

EDIT: I also tested what I can do with my new humble rank. As a result I was able to accept challenges with it, alas for friendly games only. I.e. with brand new account I was able to take a friendly game for 20-22 k players. The only missing part is the ranked games - apparently when challenges are created they use players rank calculated on the server. In such case the 9-level difference limit prevented me from accepting similar challenges from 10k (humble 13k) player.

I’ve created a github issue, where the technical part of discussion can be run a bit more in depth.


Another approach:

  1. Display all newcomers as unrated.
  2. Unrated players creating custom games can freely restrict rank, so unrated beginners can play established beginners and unrated dan players can play established dan players.
  3. Established players creating custom games can choose whether to allow unrated players, so those who want to welcome newcomers can do so and those who want to avoid newcomers can also do so.
  4. A game between an established player and an unrated player should have minimal impact on the rating of the established player, regardless of result.


Glicko-2 based systems generally don’t need this kind of feature because volatily is already built into rating mechanism. The whole purpose of Glicko-2 is to have a simple, self contained, quickly converging rating, that doesn’t require special treatment. Any proposition creating a separate category of newcomers may make the original problems worse, by frustrating them even more with inability to find opponents.

With that in mind you can think of 1) and 4) as already being part of the rating system. The other two ideas are not necessarily alternatives, but more of extensions to other propositions.

  1. Good idea, which might help quick progression towards desired rank. Not sure how to express it without introducing the category of newcomers.
  2. Probably a bad idea, as it complicates the system and stigmatises newcomers. The challenges can already be limited by rank.


I agree with Nightstalker, to expand on what they wrote a little: New players (both new to the game &/or new to OGS) would like to play [other] new players (or experienced players of their rank) [and] not get slaughtered (or get an opponent way below their rank / boring game); and not everyone will take into account how the ranking system works (heck, new players probably won’t even know how it works). Not having an option to give a preset of your rank is a very poor approach to new users (some of who will be absolute beginners and some who will be experienced players coming from KGS, other Go servers, &/or playing IRL.


Previous comments mentioned established players cancelling games upon realising that their opponent is a newcomer.

From the perspective of a newcomer, waiting a bit longer for a game is better than getting a game only for the opponent to cancel with an unfriendly message.


Agreed. People indicating they don’t want to play newcomers (hello, KGS) is better than them not being able to and then cancelling the game.


Well, that’s why I don’t play on KGS. Finding first few game especially in the dan range takes forever. I’ve had few attempts, but eventually gave up.


Newcomers should be automatically paired with newcomers. Instant game, no unfriendly backlash or cancellations, immediately start working toward a true rating.


Being a newcomer does not equal being a beginner unfortunately. Otherwise it would be easy…


Didn’t say it did. Doesn’t matter if beginner or not, theyre all 13 kyu.


There are several reasons why pairing newcomers with established players (specifically, those who are welcoming and similar in strength) will give them a better experience than pairing them with fellow newcomers:

  1. An established opponent can offer guidance on site features that a fellow newcomer would not.
  2. In correspondence games, a fellow newcomer is more likely to time out after only a few moves.


That’s only true if the relationship is voluntary on the part of both parties.

“Automatically” flinging beginners with experienced people who have no interest or talent for teaching, no patience with beginners, is a recipe for disaster.

You put :

in parentheses, but this is critical for your proposal. What you haven’t outlined is how this would be achieved in practice. Right now it’s achieved by interested beginners asking for a teacher and instantly getting half a dozen in the forum.

However, I think most importantly, this concept of “pairing up with a nice welcoming person” is completely separate from the issue of “what the starting rank should be and how that is managed”. Which is the topic of this thread. Because it’s such an important topic (IMO) it would be better not to derail it with side issues - they could have their own thread.


Closing this thread temporarily in order not to waste your precious collective energies by letting you run in circles—this is an old and long thread, and people tend to jump in w/o reading the whole thread, thus running danger to reinvent the wheel, so to speak.