2021 Rating and rank adjustments

[solarwinds]

Yay! I always dreamt of going up 6 stones when I slept

[Atorrante]

Wrong thread

You know you are a Go addict when …

I’ll start. You know your are a Go addict when you see Go shapes and joseki in tile mosaics.

[Atorrante]

@anoek

Because of #2, we removed our sliding windowed rating system, which means things will be a bit more intuitive now - if you win, your rating will go up, if you lose it’ll go down.

Good news! Never again cry baby topics like “I won my game and my rating went down, I want to speak to the manager”.

This puts a smile on my face. Although they probably will find something else to complain about. Probably “I won a game and my rank didn’t go up enough”

[meili_yinhua]

@Atorrante

This puts a smile on my face. Although they probably will find something else to complain about. Probably “I won a game and my rank didn’t go up enough”

naw, I’ll just complain about how I lost a game and the ranking system was not capable of comprehending how my true skill went from SDK to 30kyu over the course of a game

[DVbS78rkR7NVe]

The boost does sound huge, even with the asterisk of new rating not necessarily being the same as old ones. And that’s a bit annoying. As you said, it’s underserved, we all know that we didn’t really reach these ranks, some mathematical nonsense happened behind the doors and we all are dans now. If we’ll be pushed over into dans, that is.

Imagine being asked by a stuck 2k what did you do to reach dan, haha.

It can be really deflating. Especially when you push to improve and gauge the progress by your rank, and then all ranks are wiped, your progress is wiped. It can be difficult. It was like this last time for me, and even though the ranks adjusted back, I did quit go for like half a year. But I’m all better now. Let us remember. Ranks don’t matter. They don’t show your worth. It’s just a made up number. Recalculation doesn’t change who you are. Recalculation doesn’t change your go strength.

[gennan]

@claire_yang

@anoek

From the distribution, looks like the majority would be bunched in the middle around 10k?

I made similar chart for EGF ratings of European tournament players (before the coming EGF rating system update):

You can see the 20k “pergatory” that the EGF rating system has had for 25 years.

The shape is a bit different from the OGS distribution. It’s less “bell”-like and it has its peak around 7k (that peak will shift up to around 5k after the coming update).

It looks like the OGS rating system update will move the peak from about 18k to about 10k. That is still lower than the EGF peak, but that makes sense. I expect a typical European tournament player to be stronger than a typical OGS player.

[DVbS78rkR7NVe]

Let’s play go games in this last day of old ranking system!

[claire_yang]

EGF has amateur 8d? Who are they?

Amateur 8d is an honorary title here, only for those like won international championships. Is there any special requirement to get to 8d in EGF or just rating based?

[Animiral]

@anoek

Now mind you, I tested this code once upon a time and it was correct once upon a time. However, at some point I accidentally deleted a character (which is sort of easy with my editor of choice, vim) and because nothing exploded and generally speaking, looking at one game change, the results aren’t that crazy, I didn’t notice it. For years. That bug had been with us since we introduced Glicko2.

Thanks for the detailed disclosure!
Do you have automated tests?
Do you run them as regression tests before release?

For a project of OGS’ scale, they are well worth considering.

[BHydden]

IIRC, the tests are on a test server which is meant to be identical to production, but somehow this bug came in after testing and before push to production.

[MooToYouToo]

Really great news!
I’d be interested to know what the next priority is for OGS after this update settles.

[Chinitsu]

@MooToYouToo

Really great news!
I’d be interested to know what the next priority is for OGS after this update settles.

Adding Automation Test

[gennan]

@claire_yang

EGF has amateur 8d? Who are they?

Amateur 8d is an honorary title here, only for those like won international championships. Is there any special requirement to get to 8d in EGF or just rating based?

8d EGF is not an honorary title.
These players are very strong Korean amateurs who were insei once, but never made it to pro in Korea (competition to become pro was pretty stiff in Korea 10-20 years ago).

When they moved to Europe, they started at 7d, but they generally beat almost all native top European players (including those who got a pro rank in the Far East or Europe), so their rating gradually increased to 8d EGF over the years.

See for example Young-Sam Kim | Player card | E.G.D. - European Go Database and List of Players EURO**

They also tend to teach in various go schools and go camps across Europe.

[Sallmard]

Wow! Using VIM! The kind of developer I aspire to be

Sorry - a lot of info to get through: will the historic graphs change or just the forward? I’m still the same rating as yesterday so will wait a bit to see if it beds down

[anoek]

@gennan

In the range of 30k - 5k, it is quite a close match (30-70 Elo per rank).
So the new OGS formula will make OGS kyu ranks as soft/tough as EGF kyu ranks and I expect that OGS kyu ranks will align much better with EGF kyu ranks. Definitely good news!

In the high dan range, the EGF formula gets tougher than the OGS formula.

So over time, EGF mids dans may get high dan ranks on OGS.
It may turn out that OGS high dans will be comparable with high dans on asian servers.
But I’m not saying that this is a bad thing. In the past, EGF 6 dans could be 9d on KGS and I didn’t hear much complaints about that back then

That’s really interesting, and reaffirming that the shift in kyu ranks might indeed be the right thing to do, thanks for sharing that. I’m definitely curious to see how our new dan ranks perform - I’m not anticipating large swaths of high dans or anything, but I guess we’ll see how things look after a little bit.

[claire_yang]

@gennan

See for example E.G.D. - European Go Database | Main panel and List of Players EURO

They also tend to teach in various go schools and go camps across Europe.

Wow, you guys have graphs and detailed match histories and all that. We only have a list of dan players, and the records are everywhere in each tournament website or news, etc, not in a database we can look up. I don’t think there is even a list of kyu rank players that can be tallied, with like 40 to 50 local associations. They probably have their own SDK player lists who participated in local tournaments, but some might still be paper documents.

BTW, How do you make a bar chart out of all these? Doesn’t look like it can be done by hand.

[anoek]

@gennan

I expect a typical European tournament player to be stronger than a typical OGS player.

Yep, we’ll see where the dust settles, but as targeted an average EGF 1d will be something like a weak OGS 2d, and an average OGS 1d is like a weak AGA 2d. That diverges more in the lower ranks, with EGF ranks being stronger by 4-5 stones in the ddk range, although as per your previous post, it sounds like you’re bumping that range up a little bit too so it sounds like our lines might be a bit more parallel, just with EGF being a bit stronger than OGS across the board. I’m excited for your update, thanks for all the work you’ve been doing there and for putting your notes here for everyone to see

[anoek]

@Animiral

Do you have automated tests?

I do now have a test for the correctness of that code

[anoek]

@Sallmard

will the historic graphs change or just the forward? I’m still the same rating as yesterday so will wait a bit to see if it beds down

The graphs will change, and the change hasn’t happened yet, it’s scheduled for about 12 hours from now

[mlopezviedma]

For those of you who are worried about reaching dan or some rank milestone you used to have, here’s what I suggest: When your rank graph is updated, take some minutes to study it a bit. Check out where you did reach your desired goal on this new graph, check out the games you played at that time, do a tiny review of the most relevant ones. Internalize all this so you can say “this is where and when I did it”. Take this as what it is, a site bugfix and improvement. It’s all for the better.