My current overall rating, and my 9x9 rating, is 9k. For 19x19 I’m currently a 13k, but that’s likely inflated: most of the more recent matches that I’ve won have been due to my opponents resigning or timing out, even though they were significantly ahead. In my last match my opponent was 11k on the 19x19 board, however I only got a 0.5 Komi.
Is there a way to restrict the matchmaking to 19x19 ratings instead of overall rating? I’m getting slaughtered in my 19x19 matches.
There doesn’t seem to be. It’s quite annoying, it seems like the only option is to just keep playing 19x19 until my overall rating drops down close to my 19x19 rating. But this seems like sandbagging, and then I could be too strong when I play 9x9 matches.
I guess I could make a separate account for 19x19, but I just signed up as a supporter so I wouldn’t get any of the benefits on the new account.
It’s possible to mention this in your game description, restrict game settings to ‘no handicap’ and also restrict the rank of players who can accept your challenge in the game settings before creating the challenge.
(choose the option below the colour and handicap settings to restrict the level of players, then select the rank range you would prefer to play against)
(for example, something like ‘for 14k-12k players, I am around 13k in 19x19’, whilst restricting the challenge to 14k-12k players)
In automatch settings you can select to require no handicap and set the range of opponent ranks from yours. So you could set say 0 to -9 and no handicap and only be matched with opponents from 18k to 9k with no handicap.
If I’ve understood how settings work!
Obvs this might increase wait time for a suitable match.
I would just say that it might not take as long as you think for your 19x19 rank to “catch up” to your 9x9 one. A lot of the reading and fighting skills carry over and once you’ve learned a bit about the effect of a bigger board you’ll probably find matching on both is fine using overall rank.
There are several threads about this on the forums. I can’t remember this ins and outs but there gist is that in general overall rank is more accurate than the individual ranks which shouldn’t be taken too seriously.
…as a site supporter, you can upload your games to the server manually (to your SGF Library on OGS) and queue them for a review as well.
They needn’t necessarily have been played on your site supporter account to be analysed by the site supporter analysis features.
So one solution may be to upload your sgfs (game record files) from the other account to this supporter account for site supporter analysis of them.
Of course, another solution might be to take this as an opportunity to improve in 19x19, and as @teapoweredrobot said, it may not be as difficult as it seems to ‘catch up’ to a rank closer to your 9x9 rank once you’ve grasped some of the key concepts/global judgements which are different on 19x19 to the 9x9.
Ooooh, I assumed “Require” just forced handicaps, I didn’t realise it had additional options!
Interesting, that could actually be quite helpful.
Maybe you guys are right. When I think about it, in a few of my matches the post game analysis says I was able to keep up with my opponent for a few dozen moves, but then I just end up chasing one of their groups. I guess I’m too used to the edges of the board helping me to control them.
I guess most players on the automatch queue are using the default ±3, so increasing a boundary over 3 should have a neglectable effect, while reducing below 3 will reduce the number of possible opponents (=longer waiting time)
Yes, it’s usually much easier to run and save groups on the 19x19 compared to the 9x9 (and more difficult to force the in the same ways as in 9x9) given the wide open space, so there are more possibilities to account for in this regard.
Even though I began playing only on the 19x19 for many years, when I first began playing 9x9 recently and dedicated much time and effort to learning it, I noticed a tendency to over-focus on the local or over-attack a bit when I began playing more 19x19 again.
Though it is more difficult to fully surround groups or continue attacking them for a long time, one can judge when to continue attacking on the 19x19, when it may lead to over-attacking or over-extending for the time being, when to leave it be to take a bigger point globally (even one some distance away, which doesn’t attack directly but may also help attack a weak group in future), and also how to make profit even when a group isn’t killable.
(for example, forcing the opponent to play moves across neutral points, whilst securing solid territory by playing moves on the 3rd, 4th or 5th lines surrounding territory during the attack, or whilst gaining more of a whole board advantage through making strong shapes/influence in the attack…
or surrounding and forcing the opponent to make life small when possible - with only a few points - allowing the group to live small and sealed in rather than chasing it around into the open areas of the board, whilst building a strong outer group/influence which can impact the global balance in your favour.)
It takes perhaps a bit of experience and time to learn, but I feel one can develop a sense of the flow of this on a 19x19 after seeing what works (through game experience), or perhaps looking at some professional games and the pacing of the opening/midgame development in them.
(For example, Shusaku’s and Shuei’s games are both considered very ‘clear’ styles to learn from and play through, though from the no-komi era. But I feel looking through any number of pro games from any era within the past few centuries which pique one’s interest can help one get a better sense of the whole board and pacing decisions on 19x19. )
Unfortunately this is not possible. In my opinion that is one of the worst aspects of OGS. I know have three accounts that are logged in in three different browsers for that exact reason. They say it’s more accurate that way (I assume that’s simply because there is more data when you include all sizes), but it’s less specific. That makes the match making rank less useful for finding fair matches.
There are many profiles that play on all sizes, where the overall rank is better than any of the individual ranks. That is curious and it’d be interesting to look at the data to see why that is. However, if the correlation of ranks on the different board sizes was strong enough to warrant the use of a single unified rank, then that should not happen.
Also, it influences how players choose what to play. I know I stopped playing 13x13 on the main account, because it inflated my rank too much, so that I’d only lose games on the big board. It’s no fun. I suspect the reason for this is that the players that play a lot of 13x13 are weaker on average. I know for a fact that I’m not alone in this. Playing smaller boards tends to inflate your rank. Not because you’re better on smaller boards, but because the rating system is calibrated for a different player base.
I hope that at some point in the future, the developers take another look at this. Hopefully with someone that understands statistics. But for now, the best solution is to use multiple accounts unfortunately.
You answer your own question: because there is more data.
For example, let’s say that we have a player ranked at a uniform flat 15k across all sizes. However, they have since improved their game and they win their next three games against three 13k: one on 9x9, one on 13x13 and one on 19x19. Naturally, their rank will go up, maybe to 14k on every size. But only the overall rank takes all data into account, which is why it adapts the fastest and is the most accurate. In this case, it will be the closest to or above the 13k opponents that were beaten.
If you have the qualifications, I’m sure the OGS team welcomes your contributions. You might start by familiarizing yourself with the work that has already been done and written about on this forum. e.g. these kind of threads as a starting point.
if you do too narrow rank restriction when search a game, you will wait too long.
But if you do less narrow rank restriction, it no longer matters if your rank is inaccurate, because you will accept opponent of different rank anyway.
So problem is more little than it looks.
I think I’ve found the same phenomenon. I play only 9x9 games, but my ranking is not directly transferable to a standard 19x19 rating, since I never play 19x19 games (I like playing shorter games, both live and correspondence types). Is there a formula I can apply manually that would give me my approximate 19x19 ranking?
It might seem that way, all things being equal. But statistics is a science that can detect similarities between different systems. I would not be surprised if a formula would give an approximation of the standard 19x19 rank, given the 9x9 rank or rating.
For a start we need to fix what we compare. How long an experience on 9x9. How many games on 19x19 in the new experience?
Seems quite not obvious, too many diffferent situations.
Not saying we could not try to collect some very extreme cases (like first game on 19x19 after regular practice on 9x9 for let say a year) but even here i doubt to get some consensus on what could be the difference of level. May vary a lot in my opinion.