OGS has a new Glicko-2 based rating system!


#41

While I understand the evolution and improvement of OGS to a new ranking system, and for the record I say it is a good thing, the disclaimer that only a slight movement in your current ranking, is totally untrue. I just checked my “new” ranking and I went from 16K to 20K, I am devastated and angry. It as thought my 500 wins and hard work to even maintain a 16K means nothing! All they years playing here, and getting stomped, learning, striving to be better all has been trashed.

Should I have played better? That goes without saying, it easier to visualize SDK when you’re 16K vs 20K, it might not matter to most, but it matters to me, I hope there will be a formal process for those who dispute the rank issued by this new process, as I will be the first to protest, I know I am better than what has been handed to be, and I am willing to prove such.


#42

Hello @kissmecomix,

maybe it’s a good idea calm down and wait for a day or so until the new system has worked through all players and all games (I must admit that I myself don’t really understand the goings-on under the hood).

But I’ll mention @anoek anyway so he can maybe explain some more. I think the ranks are still changing, and there may be a few bugs to weed out.

So … please don’t despair yet, OK? :wink:


#43

@kissmecomix your go skills have not changed a bit from this ranking system change. It does not matter if your rank is 5dan or 15 kyu. You will still be able to play against people who are at the same level as you are.

If this is the case then simply playing is “proving” and the new system should pretty fast put you in to the correct place in ranking (if there are no bugs)


#44

Maybe I missed this somewhere obvious, but how do I interpret my new number vis a vis a Kyu ranking?


#45

You can ask the moderators to change you back to 16K in a few days if you wish, per the post at the top. Or play a few games and it will adjust more quickly than in the past.


#46

or, in other words, you kinda don’t :stuck_out_tongue:


#47

By looking at your profile I see you hit 15k a couple times. But since February you’ve lost over 50 games against much stronger opponents. The improved accuracy of this rank for everyone will help you find even games and then you will start climbing again.


#48

I want to ask you, why did you decide to change the Dan Ranks up to 10Dan and make them basically incomparable to the EGF Ranks?
Is it not possible to maintain the old achievementsystem (7D = 1p and 8D = highest level of achievement for an Amateur) in combination with the Glicko-2 system?
I really liked the fact, that OGS was the only Server in which the Rank had significance.


#49

I agree that 7d should be the highest you can get by playing.


#50

I too am playing someone who’s new rank is N I’ve refreshed many times but rank is still N but I know he is stronger than this. His name is paulpster


#51

“Correspondence games between two players that rank difference of more than 2 and end in a timeout will no longer be rated.”

This means that a player who only open games with players whose ranking is different by two or more and abandon all games in that it is at a disadvantage will win every games?

This means that if I’m with advantage in a game with one player whose ranking is different by two or more and he doesn’t want to lose enough losing for timeout and my effort will be wasted?


#52

Why is this?

It seems odd that correspondence is singled out in this way. What’s the rationale?

At the moment, the way this reads is “you can no longer play ranked games by correspondance with people more than 2 ranks difference” (since the loser will just time out and escape).

GaJ


#53

Also what about the case where you start a correspondence game the same rank but by the end the difference is more than 2? There aren’t even reasonable steps you can take to protect yourself against dodgers.


#54

Cant you cancel out before playing a move, without penalty? The way I read it is that only after each person has played a move then no more cancel. Which is fantastic - that’s what we lobbied for!

EDIT: OH! No, I see: Bhydden’s point is that at the end of the game, your loser opponent may have dropped to 3 ranks below you and then can escape.

Hmm - interesting corner case!

But really, it’d help to understand the motivation for the rule in the first place!


#55

Wow, nice fast work, thanks!

I should have first said: nice work with the whole thing too! My own rating went down one, which seems fair -
I seem to have retained comparison with the other beginners I’ve played (of course) and I like the tables and graphs.


#56

Alrighty, lots of pushback about this change:

Correspondence games between two players that rank difference of more than 2 and end in a timeout will no longer be rated

We’re going to redact it. The reasoning I had behind this is we used to see people ranking multiple ranks due to timeouts from their opponents, when they are strong this is amplified. I’ll ponder on some other ways of addressing the problem, preferably without such potential for abuse.


#57

Could you rate it for the loser but not the winner? So no advantage for intentional time out (except spite maybe) but no artificial inflation either. I don’t know. People just behave badly sometimes. :frowning:


#58

With Glicko, big upsets would raise the uncertainty score. Could this help mitigate unbalanced ranks due to timeouts?


#59

I think it’ll certainly correct faster, I was just hoping to do a little more. I’ve been talking to @matburt about it a bit about an idea from https://www.reddit.com/r/baduk/comments/6s70jk/discussion_ogs_new_gilcko2_rating_system/dlb5puk/ which is basically to count correspondence timeouts if someone is just ignoring a game, but if they haven’t logged on in awhile, then annul the game.


#60

That might be an idea too. I’ll have to run some tests on that, during this effort I noticed the Glicko system is a bit more sensitive to asymmetric rating adjustments than the old EGF/Elo system (which actually needed them in order to combat perpetual rank deflation)