[Outdated] Improved OGS rank histogram

@BHydden prodded me to make a proper version of OGS rank histogram. Indeed histogram from 17 months ago isn’t good at all, I didn’t put much thought into it.

This time over the span of two weeks I polled almost every player who played a game within last 6 months (specifically every player who is connected through any chain of games to Sadaharu).

Real ranks:

Humble ranks:

File with players: https://gofile.io/?c=0Y0mzB


If I’m reading these graphs correctly, we seem to have more players who are less than 25 kyu than all of our dans combined :frowning: that doesn’t seem ideal…


It’s a lot easier to be less than 25k than it is to be a dan…


What is humble rank please?


@teapoweredrobot It’s this Humble rating and rank formula

I didn’t think it was THAT easy to be below 25k :thinking:


Perhaps you know that our rank has a value and an uncertainty, like my rank is “11.4k ± 2.4”

This means that as far as OGS knows, my rank could be as good as 9.2k or as bad as 13.6k

As we play more games, the uncertainty gets less - OGS gets more and more sure about our rank the more we play. I haven’t been playing much (moderating and coding) so my uncertainty is quite high.

Simply thinking, “Humble rank” is the lower bound of our rank. So my humble rank is 13.6k

“But so what!?” you say.

Well, humble rank is used for match making, when our uncertainty is high.

Our uncertainty is high when we first join.

When we first join, our rank is approximately 13k +/- 6k. IE OGS has no real idea what our rank is, but it guesses somewhere in the middle.

You can see that when we join our humble rank is about 19k.

This means that new players are matched with about 19k people instead of 13k people.

This actually solves a big problem we used to have, which is that new players would be matched with 13k players, and both of them would hate that. The new player would hate being matched with someone so experienced, and the 13k would hate getting games with new players all the time.

That’s why, when our uncertainty is high, the matching system uses our humble rank.


I think that based on the graph above, it is self-evident that it is.

Maybe take a look at a few new player’s rank graph to convince yourself how this happens…

Depending on the legitimacy of those <25k rankings, there may be justification for opening the rankings to a range of 35k-9d as has previously been suggested elsewhere.


I’m not doubting the evidence, simply stating my shock at how off my internal compass was on this matter.


I had a similar thought Kosh


This is the basis for many people’s complaint that they get “stuck” at 25k for ages - the reality being they dip well down below 25k, but we show 25k the whole time, so it takes a while before they surface again above 25k and see progress.

Which is the point you and Kosh are subsequently making.


Yeah I knew it was a problem I just didn’t realise how extensive it was…

1 Like

Thank you @Eugene for anticipating my follow up question!

And on <25k, in the British youth club “system”, 35k is where you start when you can finish and score a game (9x9 or 13x13) without assistance. Leaving aside trickier/special situations which comes up even more rarely for absolute beginners, it seems a good starting point for getting on the kyu scale.


“suggested elsewhere” is code for I couldn’t find it!

But now I have. Do you have any idea how many forum posts include the word rank?! :rofl:


People are more concerned about their rank than the size of their p**is.
Thankfully spammers didn’t notice yet, otherwise we would receive tons of mails saying “enlarge your rank”! :crazy_face:


I don’t think there is much point in ranks lower than 25k. At the moment, I’ve the impression that if a player recognizes self atari at a regular basis and stops playing them, they move to 24k and above.

Additionally, ranked games are only allowed in a range of ±9 ranks (auto match uses ±3 ranks). To lock weak players in with other players witch don’t recognize the basics of the game doesn’t do any good. They wouldn’t have even a chance to see how the game is played.


True story. Can confirm. :joy:

Big +1 on this. I was stuck in that 25k pit for a while and it felt really discouraging to not see any change in my rank after multiple wins.

I’d imagine this large number of sub-25k users also is a result of new players joining, playing a few games, saying “I don’t get it”, and moving on to something else.


On top of this, I have a friend who is well below 25k and it’s pretty hard to watch them get SLAUGHTERED by players supposedly the same rank


This was true for me as well. It would have been really nice to be able to see some progress. I was stuck there for way too long and it was extremely discouraging.


You could get around this by ensuring that anyone below 25K only gets paired with players who are a bare minimum of 25K, or whatever K you feel fits. This would ensure they always play someone more advanced than them.


I was really hopeful the large numbers were proof that OGS was flourishing with newcomers. But your interpretation seems much more likely.

:thinking: Is there any way to check the 25K account range against account activity? Maybe check how many of those accounts have played a minimum of 5 games and have been active within the last two to three weeks. That would be a much better representation of what the 25K pool really represents.