Time to make dan ranks weaker?

Currently I’m have provisional status here.On Tygem I am 3 dan.On KGS I am 1 dan.I do agree that the ranks here seem a bit off.

Good things to know. The plan has been to reevaluate our rating and ranking system after the updates we put in a couple of months ago to see how things are holding up. Given that we didn’t bump dan ranks up during that update, I suppose it’s not too surprising that they’re still depressed. I’ll certainly keep this in mind when we review things again (which will likely be fairly soon after the 4.2 release).

7 Likes

I agree. It is important that ranks here are roughly comparable to other servers if you want people to come here. The rank intervention early this year was hugely important in my opinion. We may need another.

I have the same feeling but I do not play enough to compare both servers. I used to be 3d on KGS. 2 friends of me also have a similar rank on KGS and are about 1k/1d here.

Was more for information than anything else as I don’t really care about my online rank as long as rank difference between 2 players is correct :- )

I’ve often wondered how the rankings are computed, what algorithm is used … I wonder specifically in terms of whether it is better or not to play a game out rather than resigning. Score should obviously be a factor. But in an early resignation the score estimator should be severely off, so would playing the game out, at least to endgame, be better for your ranking?

OGS uses the Elo system as described here: http://en.wikipedia.org/wiki/Go_ranks_and_ratings#Elo-like_rating_systems_as_used_in_Go
The final score in a game has no effect on the rating change, so playing out a lost game won’t give you more rating points than resigning.

1 Like

Score and manner of victory (i.e., timing-out vs resigning vs final score) are not factors, and they should not be factors in computing ratings.

Here are some reasons why they should not be factors:

  1. Incorporating score as a factor would change the game significantly. A player winning by a solid margin would be discouraged from “playing safe” while ceding some of their margin. A player losing by a slim margin might even be discouraged from playing bold moves that could potentially reverse the game given the risk of further widening the margin.
  2. Incorporating these factors would be complicated and controversial. Difficult, and likely arbitrary, design decisions would have to be made. Rules like “10 points equals one rank” are very rough (and bad) heuristics at best.
  3. Worse of all, depending on exactly how this was implemented, players could potentially abuse the system in various ways, e.g., (potentially) resigning to avoid a large margin of defeat, lengthening a hopelessly lost game, timing-out to escape a worse rating impact, etc.
8 Likes

Excellent explanation. Apparently the only two things that matter are whether you win and what your relative ranks are.

Well, I was 7 or 8 dan on OGS and 3 or 4 dan on KGS (and 6/7d Tygem), but that was under the old OGS rating system before the Nova merge and I did play better here…

Right now, I am 5 kyu KGS and 11 kyu OGS. I’ve played many games at OGS and I think I am a bit underated right now, but based on previous matches I’ve played, I think my level at OGS should be at most 8kyu. Based on KGS games, I think 5kyu / 6 kyu seems credible. So, I think this difference exists even in the SDK ranks.

2 Likes

I would agree with elsantodel90. I’m 7k on KGS and I find my current 11k level on OGS not easy to maintain. That being said, I don’t think that the differences between servers is such an issue. The only important thing is to have a relevant scale between players so handicap can be accurately set.

the stats at http://www.ootakamoku.com/ support the idea that different servers have different biases in their ranking. I played 12 positions, and it tells me, “By my estimate your rank is between 5.7 Kyu and 5.6 Kyu.”

then… – There’s an “I want to…” link bottom-right, and then, “…get more info about my rank.” --> KGS : 4.7 Kyu - 4.6 Kyu. OGS : 12.0 Kyu - 12.0 Kyu. Tygem : 1.9 Dan - 1.9 Dan. … lol …[ it’s kinda awesome that OGS is the best! ]

I’d assumed those stats were related to the skills of the players on the server … if I played on a server has mostly weak players, my kyu would improve! :wink:
… just a thought, maybe a natural effect of having lots of awesome players here, that is causing my score to be lower on OGS :blush:

Ootakamoku.com ranking algorithm is broken I think. Do not trust what it says.

3 Likes

As it ranks me 5k (I wish I were !), I can say it’s totally broken !

I have found that ALL sites that do tests and then try to approximate my rank are about ten stones off (to the positive, i.e. they feed the illusion that one is stronger).

3 Likes

I disagree.

First of all, Tygem and WBaduk have weak dan ratings. I am 7 dan on Tygem most of the time, and I really should not be 7 dan anywhere in the world. Their ratings are weak (at least in the dan levels) and this causes a wide range of strength within the 9d rank. I don’t think it’s a bad thing to have ratings here worth more in strength. I think it’s a great thing.

Why does OGS have to copy after other Go server ratings? KGS and Tygem are not in sync at all, not even a little bit. The rating system itself is completely different, and the dan strengths are off by like two whole stones.

OGS is going to be better than KGS at this rate! Maybe it already is! (I haven’t experienced a lot of bugs or anything yet, myself). Why should OGS tailor itself to match worse Go servers? OGS is its own thing and I think harsh ratings make the ratings themselves worth more and mean more. It means that 7d+ players will really be quite strong, and that’s a good thing.

Personally, I would sooner wish for a depression than a dump in the ratings :slight_smile:

2 Likes

There are some reasons written in the first post:

"
Dan ranks seem pretty hard at this time and it makes it seem like
there are less strong players than there really is. There is no real
point to having the strongest lower dan ranks of pretty much all servers
right now.

In fact, it might be hurting the server since it creates a lack of
2-3 dan players, which makes it harder for newcomers in that range to
find games, and stay on the server."

Oh and score should of course NOT be a factor in the effects of a game on rating.
Are you kidding me? Any dan-level players at the very least should understand that the margin of victory means nothing. Many players play a style that frequently captures dragons, while some others like to drag out the game and win by 0.5. When we are leading, we may allow our opponents to catch up a little by playing defensively but not letting them catch up completely (I don’t do this though, I don’t think it’s good to turtle up ;p).

One dead group can make an enormous difference in the score. Also, a 5d could play a 2k and only win by 15 points. Sure, the 5d could have played more aggressively and forced a battle of fighting skills and widened the score, but that would mean they have to play differently out of worry about rating. That would make no sense at all. If the weaker player plays very well and very solidly, there’s sometimes no reason why a strong player would win by a lot, because winning by a few points is enough. The fact that the 5d would win 10/10 games is what’s important, not the margin of victory.

At a local Go club there are a few players somewhat close to my level (and a couple that are totally out of my league, actually), and there’s 1-2 of them whom I play often and who somewhat often only lose to me by 4-8 points. However, these individuals lose to me close to 20/20 games we play. So it doesn’t matter how much you beat them by as long as you can keep winning! All of those games could be half-point victories, but as long as it happens consistently, I am a lot stronger than them. If I’m winning 19/20 games or better, than I’m most likely 1.5 or more stones stronger than them (don’t quote me on that guesstimate. Maybe it’s 2.5?).

Harsh dan ranks defend the honor of Sandbaggistan.

3 Likes

Well, the sandbaggers’ ranks would be bumped up or down the same way everyone else’s would, so it wouldn’t change the fact that we live in Sandbagistan.