2020 Rating and rank tweaks and analysis

I rarely play anything besides 19x19, and usually avoid ranked 13x13 and 9x9 games especially much, and I think that new players should be encouraged to jump into 19x19 as soon as they’re invested enough in the game to not quit, but I don’t see why players who enjoy the smaller board sizes shouldn’t be able to play them as well.

PS: I’d enjoy ranked 21x21 and unranked 31x31 as well, but I know that the first will never happen, and the second probably won’t.

4 Likes

If this was the case until 20k for example, I would just have to sandbag so I could play ranked games. Or I would be forced to have really good and enjoyable games with people annulled which I find even worse.

Edit: I don’t like unranked games much and find less motivation to play them. You may or may not be able to tell that I don’t like them as much from what I said. Variations are fine unranked, but if I am playing a normal game of 13x13 or 9x9(what I usually do), I like it being ranked.

5 Likes

First of all grats for fifty likes.
.Also a account I forgot the name of (probably by too much Forum or Brawl Stars)Has a high SDK (1-5k)rating on OGS and a low DDK rating in real life just because of winning a lot against bots(like I do expect I lose a lot too).So it seems that this is also going to be fixed so I could see a big rating drop for him or her because it probably takes like fifteen games to get 0.5 kyu.And then the person played against a 7-8kyu for like ten times in the past few games so the 7-8 kyu is 1D actually in the middle he did win against another 7-8 SDK

3 Likes

Frankly, as a refugee from the chess world, I have been ignoring the K, and paying attention to the Glicko. I have been rather puzzled why my K doesn’t change (either up or down, now that I know it’s capped at 25, well, now I know). IIRC, I’ve not played any handicap games on this site (at least in stones, perhaps in komi). It seems to me that komi handicapping would let us practice “real” go without destroying anybody’s idea of opening theory.
On the subject of one rating for all board sizes, in chess it’s common to have a blitz, fast, “normal”, and correspondence rating. It might be useful to do the same for board sizes, but (because?) even I have learned that the 9x9 is not the same game as bigger board sizes. It would be a nightmare, however, to track X time controls x Y board sizes.

6 Likes

Anyways, I love OGS!:smiling_face_with_three_hearts:

Cheers to OGS!

Thank you @anoek for maintaining this site, for interacting with the community, and letting us enjoy playing go in the best way!

13 Likes

Did you note that anoek explicitly mentioned that this is being considered?

8 Likes

Thanks for all this work. Your discussion seems to focus on internal accuracy. Are you doing anything to align OGS ranks with external ranks (AGA, EGF, IGS, etc., etc.)?

5 Likes

Well, keep in mind those ranks are not aligned themselves so there’s no simple alignment we could do.

It is probably about time to do another poll to build an up to date translation table though.

13 Likes

This is the first review of anything that I’ve seen that explains the details. I love being able to read about the results/findings of tests, so that I know ‘all the bases have been covered’. I’m tired of so many companies just spitting out things like "we don’t need to make any more changes’ or ‘due to some findings further tests may be of need’.

Thanks for the update!

10 Likes

probably after giving it some time to settle, I don’t reckon all of the adjustments will be neatly accurate

3 Likes

Thank you @anoek, both for the detailed statistics, and for the notification about this forum thread; what follows are my thoughts, with the warning that they are those of an inveterate gambler, uncurable drunkard, accursed sandbagger, and reluctant student of mathematics who has not yet been kicked out of university:

  • Predictions that do not cost anything to the predictor, are worth their weight in osmium.
  • Conversation between black and white influences the outcome, so the statistics might look quite different when split according to loquacity.
  • Perhaps it should be encouraged to read the profile of the prospective opponent, including a glance at the statistics, before accepting a challenge.
4 Likes

Yes, I noticed. Trying to put some pressure. :wink:

4 Likes

tl;dr: How about a separate ranking for “proper” 19X19 normal time games?

Having chatted to other players and seen what other players and viewers on certain twitch streams are saying, there seems to be a kind of consensus that the “real” game is 19X19, normal time.
the smaller boards are really for beginners, or if you don’t have much time.
Sure, play these, but playing these should not mess up your “real” ranking.

Similarly with the other timings.
Blitz games give no time to consider the game more deeply and are more of a reaction test.
With correspondence games, I can spend (say) five minutes working out how the game is going and what my strategy will be. By the time my opponent makes a move two days later, I have forgotten. I end up with a list of ten games and play them by going: next, click, next, click, next, click and develop no deeper understanding of any of them.

In a rated tournament in meat world, you would generally play normal time (obviously not correspondence) and on a 19X19 board. Sure players may play smaller boards or different timings on the side, but they can do so without it ruining their ranking.

I know you display separate ratings for different sizes/timings, but how about a separate ranking for a “proper” game?

When I feel like a good “proper” game of go, I would like to be matched (and handicapped) against an opponent of the right standard for that game. I’m not interested in whether my opponent has won countless 9X9 blitz games.

Love OGS. Will be subscribing very shortly.
-Oikake

5 Likes

Quick question:
Is the handicap taken into account to determine how a game affects your rating?

If I (17k) play a 25k player, and they get a handicap of 8, and I lose, does that hurt my rating terribly, or only as much as if I had lost to another 17k player?
-Oikake

2 Likes

Yes, handicap is taken into account.

3 Likes

Bravo. I really love OGS, and posts like this make my heart swell. This really is the best go server, I don’t understand how the others keep their user base. Thanks so much for this.

8 Likes

implementing ranks beyond 25k would be nice for beginners. Even if it’s just a decorative rank maybe and not actually “used”? Just so guys like me have something to look at in the first days and weeks of learning.

9 Likes

The question of 9x9 and 13x13 inclusion into OGS rank should be a marketing decision.

I had been opposing the inclusion from the point of view that it will result in repelling new dan players, who are used to ranks being based on 19x19 only, and will result in OGS gaining the reputation that it’s a server for beginners only.
However, the recent growth in userbase made me re-think this position, because at least a part of the growth seems to have come from the very reputation that OGS is beginner-friendly.

For this, I feel we, the long time OGS users, need to pay close attention to Oikake (g-g-g-g-g-g-Go)'s post above, because of his use of the words ‘good’, ‘proper’ and ‘right’ being the same as what I hear when I talk to non-OGS Go players. In my opinion, his view is pretty close to the norm in the Go world outside of OGS.

So, please make a conscious ‘marketing’ decision on this to be unique in the world, so that we’ll promote the beginner-friendliness at the cost of unfriendliness to outside dan players. I believe the 9x9 and 13x13 ratings being statistically accurate or reasonable, is secondary to the importance of this conscious decision for the future of OGS.

Of course OGS could continue being beginner-friendly by further developing the functions for beginners, even after making a conscious decision not to include 9x9 and 13x13 ratings (which could still be posted/calculated/used) in the calculation of “OGS Rank”. What do you think?

7 Likes

My opinion doesn’t matter at all, of course, but that won’t stop me!

In competition it’s important to be different. There’s no reason to copy foxy because we already have original foxy. It’s important to provide something of value that other entities don’t offer. And correspondence and different board sizes are a few of the features that make OGS stand out.

In my opinion excluding these games from ranks is almost equal to removing them. And if we assume that then OGS will be just another server. Do you think all these foxy players are going to switch to OGS when they hear we changed the ranking? I don’t believe so. Most likely they’re already content playing wherever they’re playing. And we don’t gain much by trying to outfox the foxy, we’ll end up just the worse version of it (with smaller player base).

In fact in Russian groups the most common argument against OGS is “why do we need OGS if we already have KGS and it’s working perfectly fine”. No matter what we do here, there’s no way to break that argument.

I understand that the Japanese perspective is more traditional but it’s interesting because this issue seems so unimportant. I play all kinds of board sizes and I never really got anything unexpected to think that the opponent’s rank is ruined by other game types.


Also, I bet lots of dans don’t want to play on OGS because with OGS ranking they’re gonna be kyus here :stuck_out_tongue:


I want to remind that we already calculate separate 19x19 live ratings. The only question is what and how to show them.

11 Likes

So, we do of course keep track of those ratings, we have history for all speed size combinations. The question I was trying to answer, and I believe I have, is which is the best one to use for computing handicaps between two players? And I believe the data pretty well indicates that using the overall rank produces the best results. I’ve updated the original post to include data cross cut on speed too, so it looks to me like yes, you should still use the overall rank when computing blitz vs live vs correspondence.

10 Likes