All new players should be ranked via bot

Then I apologise for misrepresenting your viewpoint. However I do feel that since I am primarily the only person suggesting that there could be any downside to “just letting beginners pick a beginner rank” that it is hard to interpret

in any other context, given the current direction of the discussion. [Edit: my point would be you’ve given no context to this comment at all.]

Indeed, I’m sure it has come up before and will continue to.

However I think @flovos point elsewhere is relevant

and to reiterate - why should one change the rating system (allow a new or multiple entry points) if one just wants or might be happy with a different matchmaking system for beginners?

I don’t see how you can just have a different matchmaking.
If both are new players that works but what happen between a confirmed 25k after a dozen of games and a “provisional 6k” beginner? 19 stones handicap game?


At the moment no handicap. Automatic handicap is reduced for player with high uncertainty in their rating.


I have no idea how the rating system works but statements like this did give me a sense that

In that managing the rating system is an effort for OGS rather than players, who in broad terms don’t care about the rating systems (some obvs care about their rating but that’s not the same thing)

In terms of downsides, the question is more that are the new downsides worse than the existing ones.

It’s almost analogous to this^ - preferring hypothetical downsides to existing ones

And actually on that point, it’s fairly normal to prefer potential customers than existing ones because the latter are generally more tolerant then the former. I.e. there is a higher bar to get new players to join than to get existing players to stay.

I should say that I’m not sure what the best answer is. I understand there will be problems either way. And there’s the issue of deliberate misuse the think about. My proposal for consideration would be to have some kind of humble rank that can be set at different points (25k, 10k, 3k) and use that as a shadow system until the certainty is established.
Plus cap bots ranking range as I can’t believe that they should vary that much in practice.

These are my own not very thought through thoughts, not an ogs view or one backed with any expertise on the rating system or anything really!


What does humble rank mean?

1 Like

Sorry, “humble rank” is a term applied to the rank at the bottom of the uncertainty range of ? accounts.

So previously a new account on OGS would be I think 13k+/-6 so the humble rank would be 19k. The humble rank was used for matchmaking, handicap etc. So it meant beginners were better paired with beginners.

After the ranking update new accounts are now 6k+/- something but anyway the humble rank system didn’t seem to work. So beginners were now paired with 6k opponents, who might resign in frustration and send the new account higher up the rankings!
I think the humble rank is working again now but I’m not sure what it is for new accounts.

I should have said “Shadow rank” or “provisional rank” or something else clearer.


And statements like this give me the sense that

OGS =/= shinuito.

It’s like if I read

and then concluded “ @teapoweredrobot is a moderator so OGS must hate their current player base”.

I think at the moment the rating system seems like it works fine. There might be some debate on stability of rating when players play thousands of games either a day or a month, but other than that I feel like people end up with the right rank after not too many games. It’s the method of obtaining a 25kyu rank people are objecting to, that is having to lose games in order for the system to see you as below a particular rank. Of course it might mean losing more games if one doesn’t or can’t challenge existing 25kyus.

However what people are suggesting could require much more maintence and while you say

I don’t believe that the hypothetical downsides are nonsense. For example if you look at old posts when the system used Elo/Gor style ratings

and that required solutions like

where there’s more fiddling needed to be done. Note that dropping a beginner in the middle of the rating system in Elo would likely be much worse than in Glicko, since there isn’t a concept of high deviation etc where ratings can adjust quickly as one can see in the post about swapping to Glicko.

and (if players/moderators don’t adjust the rating back then)


I believe that falls under

to which I said

and yet all I see are responses saying


which I hope you can appreciate is a bit frustrating.


Here is an idea that I think may be new, although it is so simple, I would be surprised if it hasn’t been suggested before.

Why not automatch “?” players only against other “?” players? Can’t that be coded as a special rule? It seems like that could be done without any change to the underlying ranking system. The pseudo-rank of 6k could remain as a baseline for the system to calculate people’s ranks, but it would play no role in automatching. This would give beginners their best chance of a fair game, and it would certainly relieve them of the discouragement they must now feel because of a string of crushing losses from much stronger players.

This idea would also have the advantage of greatly reducing the number of games that much stronger players must play against beginners (due to the fact that the rank-range restriction does not exclude “?” players), which has been the source of another whole stream of complaints calling for changes to the rank-range restriction.

This idea might not have worked in the pre-COVID days, when OGS got about 250 new registrations a day. However, when I stopped moderating, almost a year ago, new registrations were still at about 900 per day (down from a high, in my observation, of about 1,200). Since many players will remain unranked for more than their first day, I think the pool should be sufficient for automatching.

This is not a perfect solution. I don’t think a perfect solution exists. A significant number of “?” players are experienced players new to OGS or alt sandbaggers. However, such under-the-radar players would exist in any system. If it is technically possible to implement this idea, it would eliminate two problems without any change or upset in the ranking system.


I think the main issue of pairing [?] rank players together is that you really need currently rated players to set the scale and help rate new players.

If [?] players only played each other, the best of the bunch could end up at 2kyu and the weakest at like 16kyu (since they have say an underlying rank of 6kyu) and then it would need to be further corrected by playing against players with a real rank. (The numbers sound a bit arbitrary but that’s basically what happened when I took 5 unranked accounts on beta and got them a rating against each other.)

I’m somewhat optimistic that if the humble rank worked for matchmaking, and a new player lost to a 15kyu, they could have have humble rating of 20kyu in like 1 or 2 games. The underlying rank might be higher, but the point would be that they could play the lowest rated players quite quickly.

Obviously if the ± 9 ranks limit was lifted, they could just play the lowest ranks immediately.

1 Like

We actually do do this, but there’s not always another ? rank available. We basically have a system where we wait a short while for a “perfect match up”, then expand our reach and wait for a litle longer for a less perfect but good match up, then expand our reach, etc… until we reach the bounds setup by the user.


Thank you for your reply. I’m astonished that there are not enough “?” players available.


Well, we only feel the “problems”, when everything works according to plan, people just play a good game of Go :slight_smile:

The other thing is that what @shinuito pointed out is definitely true, when you have two 25k ? come in and play eachother, one is going to get ranked down, one is going to get ranked up, and then we continue to “feel” that one that got ranked up as a problem until they get ranked down properly.

I think it’s better than nothing, but the whole experience does need improving, for the sake of everyone. I don’t think there’s an easy fix though. You certainly can’t “just set ranks”, that was horrible when we did it for lots of reasons. I’m not a fan of forcing folks to play bots, though encouraging them to if they are a beginner I think would be good, the problem there is coming up with a good beginner bot and possibly adding some tips to help teach them while they are playing kind of thing, as you want to help set them up for a win - people hate losing against computers, especially these days. So, for that reason I don’t think just blindly throwing beginners at our current bots with our current experience is a good idea either. At least with humans there’s a chance the human can help, and the expectation from the beginner that they will probably lose against a human who has played more than them.

Anyways, it’s not a simple thing, but right now improving the new user experience, as well as the experience y’all have, is top of my list, including how to ease beginners in and protect them from sdk’s and sdk’s from them until they are ready.


I have no idea if, how or should this work, but here’s a humble idea:

Since ranked challenges have rank restriction anyway, could we ask newbies

“Do you want us to restrict your opponents’ strength for a more rewarding game experience? Click YES if you are new to the game”.

And restrict their 5 first games to >15k (example) opponents.

In case I haven’t mentioned it, absolutely not forcing people to play bots. Ew.