2020 Rating and rank tweaks and analysis

Yes, the general go strength is going up, but in terms of handicap stones it isn’t all that much, I think. I don’t think Ke Jie can give Honinbo Dosaku a 3 stone handicap and I don’t think a 2020 1d EGF can give a 1990 1d EGF a 3 stone handicap (though perhaps a 2020 1d EGF can give 3 stones handicap to an 1970 1d EGF, because EGF ranks might have been closer to Japanese ranks back then).

So I don’t think this can really explain why we would need to lower the floor by as much as 10 stones over a course of 10-20 years.

I think the real cause is that the go population in the west has expanded considerably at the lower end.

From 1970 to 1990, most go players were students of the more intellectual and analytical type (typically Maths or Physics) who may have progressed from 30k to 20k so quickly that there was little need for those ranks to be taken into consideration.
But now the availability of go (online) has made it much more accessible to people who are not students living in a university city. So nowadays the go population in the West has a much more varied background and a much wider age range (children, pensionados) and many of those don’t blast through the 30k-20k range as quickly.

6 Likes

So I got to play some ranked games with the new system today. Honestly…I really am at a breaking point. If it wasn’t for the fact that I teach on OGS, I probably wouldn’t play on it.

So let me explain my experience today and my perspective.

I got deranked to 1D, that’s fine. I don’t mind working my way back up and it is kind of encouraging to get me to play as it makes me want to reach a goal. I then played a 2D and boy was he really strong. (Relative) I felt I was playing a very tough game and at the end according to ai I was down by 2.5 until I missed a simple tesuji because of dame. Oh well, that’s the kind of game I was seeking for practice anyway. I teach way too much and need a good beat down.

So I play another game. I’m limiting the ranks to +/-1 to play even games because this is my preference of rated play, even if I have to wait 5-10 minutes to get a game. I want to study the new ai openings and work on my play while getting a rank.

I play against a 1D and this game was MUCH easier. I think I only made 1 mistake (for my level) and he gained the lead for 10-15 moves before I took it back pretty easily. I think the mistake came from my bad habits I’m forming from teaching too much lol. (Playing too fast and shallow reading.) I won. Then I ranked down…

So now I am 1k and I think maybe it’s because I had a game worth more points 16 games ago. I check and I had a win against a 3D. I think, fine that’s reasonable, until I look at it and it was handicap. Which would mean that a handicap win is worth more than an even win…

A little bit of perspective now. I’ve been playing Go for 13 years and have been teaching professionally (for money) most of that time and developing my teaching skills. Having a 1k rank instead of my Fox 5D or Tygem 5D doesn’t work very strongly for me for advertising. So this literally effects my living and ability to get students.

I try not to get mad because maybe this server is harder. I try to think “what if this is my level?” Then I should just improve and study right? But when a 2D feels like a 6D Fox but a 1D feels like a 3D fox and me playing even games is worth less than if I had played handicap games AND I’m a 1k, (which I haven’t been for almost a decade in any rating system) then all of that together is incredibly frustrating.

I want to play even games against a certain reliable level and work my way up and master my opening and basics to improve. I could accept being a 1k if the opponent’s were my level but it’s different game to game and the fact that handicap is worth more (in today’s case at least) than even games, then I’m being punished for playing my level instead of playing handicap.

Since this also effect my image as a Go instructor, I feel this is an incredibly frustrating server to play ranked games on. I love teaching on OGS and the people I’ve met are great; but if it wasn’t for my dedication to teaching, I would go play on another server with more consistent ranks and a ranking system that was easier for me to understand.

I write this not to criticize OGS, but to give the perspective of a user of the server. This is just one experience among many other users, but I hope it is a valid case for reflection.

12 Likes

I feel like we get even more complaints about losing points when winning now…

3 Likes

When ogs used the Elo calculation formula identical to EGF’s except in its handling of 13x13 and 9x9 data, ogs put a lot of effort into aligning its strength scale with the rest of the world, particularly EGF and KGS. Player polls and surveys were conducted, and when misalignment appeared to have gotten too large, there were site-wide ranking adjustments.

By the time we switched to Glicko2, ogs had become one of the largest Go servers in the Western world, and the priority had naturally and gradually shifted, in my opinion, so that the ogs scale making mathematical sense within ogs in terms of “1 rank = 1 stone handicap” on the ogs data became more important.

Aside from this handicap issue, the strength to rank assignment is not science, and there’s no right or wrong about it. Unless you are talking about the inclusion of 13x13 and 9x9 performance into the rating calculation, please read Go server rank survey
(S_Alexander is not kidding in saying “dan certificates are sold”) which was a part of the reasons why the priority shifted.

If you are saying the inclusion of 13x13 and 9x9 data is “quite apparent mismatch”, we know about it as discussed between me and anoek in this discussion above.

3 Likes

yeah, I think the “flexible 15 game ratings period” where your game is calculated a whole 15 times, instead of one time-defined ratings period, is really starting to show how much of a toll it takes.

1 Like

1: Yes, I think OGS volatility of higher ranks does not appeal to stronger players.

2: I can understand that you prefer using your Fox/Tygem 5d rank to attract more students, but I don’t think Fox/Tygem dan ranks should be used as a benchmark for the Western go world, because they don’t align well with other systems of the Western world. To me, using those ranks feels a bit like false advertisement.

When looking at the table at the end of https://idex.github.io/go-rank-survey/go-survey-results,
your 5d Fox/Tygem roughly corresponds to 3.5d AGA, 3d OGS, 2.5d KGS, 2d EGF. So it would seem proper to me if your OGS rank were about 3d.

If you’re OGS rank is consistently 1k-1d, something appears to be off.
Perhaps you don’t play a lot of ranked games against mid dan opponents on OGS, making the rating calculation inaccurate by of lack of data. Or maybe it is caused by the high volatility of OGS ratings (so it may just be a transient fluctuation). But it may also be due to other OGS rating system quirks.

3 Likes

I wonder if it will take ~15 games or so for everyone’s ranks to be fully reflective of the new system. I mean, my rank seems to be completely unchanged in the new world order but others have shifted around a couple’s of ranks. I wonder if those which have shifted will “settle” somehow after a few games and in the interim will maybe get mismatched strength pairings?
Just a thought from a position of complete ignorance about how all this stuff works!

2 Likes

Can someone help me understand why I lost around 20 points from each of my last two wins (automatic handicap)?

I went 10/13 (9/12 using automatic handicap) as 1k and dropped to 2k.

That table might be out of date. I’m 5d Fox, 6d Tygem, around 2d KGS (haven’t played on it in a while) but had quite a few games against 1k/1d OGS who felt stronger than 5d Fox players - there’s no way I can give them 2-3 stones and expect to win, so 3d OGS doesn’t seem realistic.

I don’t think using Fox/Tygem ranks is false advertisement - there are a lot fewer mid dans and up who play regularly on OGS/KGS that I’d consider Fox/Tygem ranks to be more reliable from mid dan up.

Perhaps it’s because I don’t play on Fox/Tygem and I’m from Europe, so my benchmark would be EGF dan ranks (drawn from real life tournaments in Europe over almost 25 years). And real life tournament games are more serious for me than online games.

So in my perception, there is a glaring gap of 3 ranks (maybe even 4) between Fox/Tygem dan ranks and EGF dan ranks. That is a bit too much for me to consider Fox/Tygem dan ranks realistic. So YMMV.

3 Likes

I might be imagining things, but after looking at some strong&active players profiles, it kinda feels like current system is punishing those who play mainly against weaker players.

Not a big problem on itself, but for players who are either 1) high dans, 2) above ogs’s median strength and mostly playing tournament games, the system actually does punish them. Like for example, a professional player spicyspyigo has played total of 116 ranked games in june with 112 wins and 4 losses, with the effect of his rank dropping from 9d+ to 5d after this tweak.

7 Likes

Do private games show up in a player’s public game history?

1 Like

No they do not. Private games are always unranked, so they should not have any effect in the rating.

1 Like

As a side point regarding other servers, a bunch of them allow you to set your rating going into the site? Does OGS allow this? It’s been a good few years since I made an account, and I wouldn’t have even had a rating since I just picked up the game (I also don’t want to make a new account to find out).

I just wonder how this has an effect on the rating systems of different servers. When you register say, you put yourself in at either a rank you have from AGA, EGF etc, or presumably a rating you have from another go server. I know you’d expect these things to sort themselves out, but if you happened to get large influxes of players at specific ratings I wonder would it cause bunching up of ranks, and thinning out at others where people self declare?

For instance is it Fox go server that caps the max rating you can start at? I think I’ve seen this in either Yoongyoung Kim’s videos or Yeonwoo’s. Surely that causes big turbulence at say 2D or 3D or whichever it is, since you can have 8-9D professionals registering at that rank and climbing from there.

Somehow just picking your highest rating on a list of Go servers you play on, just to advertise doesn’t seem like the best judge of strength.

How do we know it isn’t the Fox ranking system that is the issue?

I did notice this today also and thought it was odd. Then again, in the last three losses I can see, they did lose in even games to a 3d in 19x19, a 4d in 13x13 and a 2d in 9x9. You might say it wouldn’t be as drastic if the other ratings weren’t included, but that’s already been discussed.

However, I also think if we didn’t have this somewhat arbitrary division in ratings that are assigned to 1kyu and 1dan, would we have as many problems or complaints? Strong players want to be a ‘dan’ everywhere. If we were only looking at ratings we wouldn’t be worried that spicyspyigo wasn’t hitting 4000 rating, we would just say there’s not enough (strong) players in the system to allow for ratings to go that high in any reasonable timeframe (at some point you’re probably gaining 1 or 0 rating points per win). BUT if you put an arbitrary 9 dan rating up there, you have this problem, that professionals are 9 dans and so shouldn’t they be 9 dan on this server too? (9dan is something less than 3000 at the moment right?)

5 Likes

That 3d is really strong, he also have wins against cornel11 and abcyuya (both high dans),
That 4d is xhu98, who is / were / should be high dan (he is almost on pro-level and 9d fox, at 4d he is only 1 stone below a real pro (which makes sense in this situation))
About that last 9x9 opponent i have no idea who he is, tbh it looks like a botter (i am really sorry if i’m wrong about it, i dont want to offend him), In any case, that game should not affect the current batch anymore, spicy has played over 15 games since then.

But the thing is, for really strong players its almost impossible to find opponents who are stronger than they are. Currently it seems that rating system is somewhat punishing those players who dont play many games against stronger opponents, they never see a positive effect more than couple of glicko-points in their rating.

When the rating system drags professionals to 5d, the 9d-amateurs (like xhu98) will drop to 4d, and regular mid-dans (like clossius) get pushed into kyus. This effect will of course eventually reach all the way down, making ogs ranks even more brutal than they were before xD

Tho, maybe this fixes itself with time when few of these 15-game batches has been played for all of us?

5 Likes

Maybe this is why I don’t progress. I thought the point of playing was the joy of the complexity of the game, the always learning thing and the analogy to so many aspects of life. Rank is just an indicator of playing strength within a given pool of players enabling matching. I don’t see it so much as the point of playing as a tool to enable good games.

14 Likes

Higher ranks certainly don’t mean the game is more enjoyable.

I had the most fun when I was around 15k-10k.

Playing “for rank” is why people bot and cheat and shit.

13 Likes

Not if the source is quoted. A Fox 6d is not a 6d, it’s a Fox 6d. An EGF 6d is not a 6d, it’s an EGF 6d. Ranks only ever have true meaning inside their own rating system.

A Fox 6d is, by definition, a very realistic Fox 6d. It is no more, no less. There is no such thing as a “true” 6 dan strength.

A little piece of me just died. I may as well stop playing, as I have long since ceased to care about my rank.

Oh, maybe I should keep playing, this sounds much more like fun!

18 Likes

I still don’t understand why I demoted from 1k to 2k with 75% win rate in 12 automatic handicap games (I won 9 straight games as 2k before promoting so the 15 games window shouldn’t have that much of an impact) T.T

@topazg Hey, good to see you again. Hope everything’s going well!

4 Likes

You too, was delighted to see that name pop up in the thread :slight_smile: