Rank question & board size

If I continue to only play 19x19 games from now on, will my 13x13 and 9x9 rank always weigh me down if I improve on the 19x19 board? As I continue to play Go, I really am more interested in the full board game, so it would be a bit annoying to have to go back to smaller boards occasionally to bring up my overall rank to my 19x19 score. Or does the ranking system only take into consideration your last 15 games, not just last 15 games of X board size?


No, you don’t have to play all of the board sizes. Your overall rank is just based on all of your games across all board sizes, but it won’t negatively affect your overall rank to stop playing a certain board size.


Ok thanks. Does that mean then if you win a game against a stronger (or weaker) rank on any board size it equally affects your overall rank?


Yep, but you can see your ranks for each size on your profile.


This is not good though. Smaller boards are quicker to play, require fewer moves, are less complicated, etc. as compared to 19x19. Therefore, smaller boards should contribute less to rating than full size games.

Arguably, similar logic could be applied to time settings.


Buy are the odds of winning/losing against weaker/stronger opponent different to a 19x19 board?

Arguably, yes. Someone (@S_Alexander I think?) was on recently saying how 19x19 ladder is locked out by strong players but anyone can get to top of 9x9 and 13x13 ladders.

For blitz and live, I see your point, but my performance in correspondence is probably the least accurate measure of my actual skill.

1 Like

I’d say there’s more noise/variability on 9x9 for sure. The rank itself is probably more or less the same.

1 Like

If one were to weight time settings differently, it seems that slow live game should have a weighting of 1.0 with faster time settings progressively less than that (like this: EGF tournament class at Sensei's Library). I guess this is because slow live is “proper” Go (in a sense) and people make mistakes because of time with faster settings so, for example, blitz is not a true representation of strength.

Correspondence also should be less than 1.0 for similar reasons (people play better because they have no time pressure at all, no need for game stamina, etc.)


I agree with you, I think 19x19 should be weighted 1.0 and the smaller board sizes weighted less. I would also make live games weighted 1.0, correspondence something like 85% and blitz like 50%.

My correspondence rating is higher as I have time to read out, but I think someone’s true rank would be in a live game under reasonable time constraints like 1m/move or something. It would be nice if there was a “standard” live time and correspondence setting that would weight rank into play accordingly.

1 Like

I don’t believe 9x9 or 13x13 are any less complicated than 19x19.

Sure the space of possible configurations or possible games are bigger, but I don’t think at a high level the games are simpler.

Endgame can be pretty hard, and fighting can be pretty fierce on 9x9 and 13x13.

Maybe for people that don’t play it regularly?


No I just think that one silly mistake can actually lose the game. On 19x19, skill has more time to make up for a blunder. Objectivrly, there are more opportunities (<80 moves vs ~150 moves).

1 Like

Objectively there’s many people that resign after one large blunder in joseki that puts them 20 points behind…

1 Like

Yeah but like it’s feasible to come back from a 20 pt blunder in the early game, especially if you are more skilled than your opponent. In 9x9, if you make a blunder in the early or midgame, chances to make it up are limited.

1 Like

This is true of course.

I don’t know if there’s any more noise in 9x9 or 13x13 though than is from Blitz in 19x19.

But I don’t know that we should be throwing arbitrary weightings on any board or time settings.

Well there is the automatch time settings which I kind of feel are standard in some sense.

This isn’t always true; lots of people don’t want to spend their whole day thinking of a response, they just read as much as they need to and answer (could be even less time than they might in a live game). Little bit less time pressure sure, but I doubt you’ll find many people using the 24hr increment to make a move like you might in Fischer or byo-yomi of 30s, to make sure they find the best response.


I’d definitely be interested to see the data (like “do upsets actually happen more often and at what frequency?”). Everything I said above was speculation, but intuitively I think it makes sense.

Define an upset though.

Supposing I’m ranked 2kyu and I mostly play 19x19 but I lose to a 10kyu who mostly plays 9x9, in an even game in 9x9, is that really an upset?

What if I’m a complete beginner at 9x9? In theory some skills are transferable of course from 19x19 to 9x9, but one huge skill doesn’t as much, and that’s tenuki-ing to play the next big point. Another is that the joseki don’t really play out the same, and the komi can really make a difference.

One can lose very easily in 9x9 endgame by 0.5 point or so, by not being used to counting or not playing perfect endgame, when there’s no chance to kill something, especially when playing as Black I feel.

Maybe what I’m saying is also speculation, although I’ve played a few 9x9 games between here, GoQuest and the badukpop app.

In any case, I’m still not sure it warrants messing with the weighting for the overall rank. There will be people that only want to play blitz, or only correspondence, or only 9x9 etc. I don’t see why their ratings should be forced to adjust slower for no reason.


Upset: higher ranked player losing to lower ranked player

Assumption: that skill on 9x9 is strongly correlated to skill on 19x19. Not a perfect assumption, but good enough that OGS uses it for the ranking system.

That scenario is somewhat irrelevant though. It’s a good example to think about, but not really related to what I am saying.

I am saying this: if we somehow know the ranks of the two players perfectly for each board size, there will still be more upsets on the 9x9 board.

I can’t prove it without data, but that is my hypothesis :slight_smile:

1 Like