A compendium of beginners getting discouraged and confused by OGS's new accounts are 12k design

When you make some histograms of other 13k games, I expect a fairly large variation in the shape of the distribution.

I suppose you could try to fit some curve to the histogram of a single game and from that calculate a rank, but I think it will have fairly large error bars.

But I will give you that it may render more information than just the win/loss results from a few games against established players.

1 Like

I’d love to see how such a classifier performs! Even classifying Beginner/Immediate/Expert correctly with ~80% accuracy would be useful.


I believe bots which analyse & estimate one’s level from games indeed exist, but I am not sure of their accuracy, especially for beginning players.

A player and friend whom I’ve known has used such a bot to estimate the level of an sdk player fairly accurately in their progression.

The bot (a fairly strong bot, capable of playing at a professional’s level) was able to provide an estimate of the player’s level after having played a game with them.

(although I don’t know the details of the bot and would have to ask them about it)


first 170 moves in each game

10 9k games:

10 1d games:

these are graphs of mistakes size, beginning from 1 point loss

average size of mistake, blue 1d, red 9k

number of mistakes:

raw graphs of 9k and 1d look surprisingly similar
average mistake size of 9k usually bigger than of 1d
number of mistakes of 9k almost always bigger than number of mistakes of 1d

there is 1 anomaly among 9k, which is like a dan in both number of mistakes and in average. Likely a sandbagger.


horizontal: number of mistakes
vertical: average size of mistakes
each circle is game


and now I can!




I’m not sure if that’s an argument for or against?

Isn’t it more likely to get players not at your rank the more you allow people to arbitrarily choose starting ranks?

When the OP says later

Yeah I just played my first game on Gopanda and won quite comfortably at 17 kyu, I guess the ranks on OGS are all over the place

it seems like they’re just wrongly ranked on pandanet, which of course winning is a better experience, but possibly from the other pandanet players perspective they too might say

Like every other game I play is against some guy that has 5 games played and just completely destroys me.

like the OP did about OGS.

It doesn’t seem like a valid critique of the rating system, though it is of course a valid critique of the new player experience, which will hopefully improve from picking a rank range.

1 Like

Im only at like 17kyu rn, was at 15kyu at one point

Wouldn’t that mean they are only matched with new players who have lost at least one game?
To protect them from stronger players, new accounts would have to start at an even stronger rank.
Or am I misunderstanding something?

So if we were to held on to every single beginner coming in, we’d have thousands of new go players on OGS each month.

So the question is what happened to them? They all seem to vanish without a splash.


Isn’t this the same in every hobby whether Go, Chess, yoga, aikido etc. A bunch of people sign up all the time, but only a small fraction stick with it, and the rest move onto something else.

The good news is that re that post from koba is that the rank restrictions were already removed so it should be possible to play bots of whichever level from the start.


A significant portion of new registrations may be fake accounts, including spam bots and other forms of chicanery.


Yeah… :<

Some of them do stay for longer and get stronger, but vast majority play a game or few and then give up. I dont have any actual stats, but i suspect that only <5% of all new go players will actually start playing regularly ://


Having promoted the game along my IRL life 5% is maybe not so bad as it looks.
If even 95% ran away, a positive first experience may bring more players around too.