KGS and its rank


We all know KGS, you may like it, you may dislike it. But we must all agree in how crazy his rating is.

Last September, KGS decided to modify its internal rating. Now, it is supposed to be more difficult to reach high levels while the lower levels are narrower; so lots of players concentrate at the 15k-1k zone. Everyone was affected by this change and players' rating grew about 5 stones by this procedure. Everyone was affected in the same way, so all your opponents also increased by 5 stones their level.

It is well-known that internet ratings are slightly higher than the real ones: as you can see KGS' ratings are always a few stones higher than AGA's or European's on the lowest levels which should be restored somewhere later to have a good equivalence on the higher ratings.

Where is that?

It's on the 9k. As you can see, a 9k KGS player varies up to 4 stones on each real scale.

 AGA   CHINA  EUROPE   JAPAN   KOREA  |  IGS  DGS  KGS
--- ----- ------ ----- ----- | --- --- ---
26K 30K 28K 25K 32Gup | BC
25K 29K 27K 24K 31Gup | BC
24K 28K 26K 23K 30Gup | BC
23K 27K 25K 22K 29Gup | BC 25K
22K 26K 24K 21K 28Gup | BC 24K
21K 25K 23K 20K 27Gup | 22K 23K 15K
20K 24K 22K 19K 26Gup | 22K 22K 15K
19K 23K 21K 18K 25Gup | 21K 21K 14K
18K 22K 20K 17K 24Gup | 21K 20K 13K
17K 21K 19K 16K 23Gup | 20K 19K 13K
16K 20K 18K 15K 22Gup | 20K 18K 12K
15K 19K 17K 14K 21Gup | 19K 17K 12K
14K 18K 16K 13K 20Gup | 18K 16K 11K
13K 17K 15K 12K 19Gup | 17K 15K 11K
12K 16K 14K 11K 18Gup | 16K 14K 10K
12K 15K 14K 11K 18Gup | 15K 14K 10K
11K 14K 13K 10K 17Gup | 14K 13K 9K
10K 13K 12K 9K 16Gup | 13K 12K 9K
9K 12K 11K 8K 15Gup | 12K 11K 9K
8K 11K 10K 7K 14Gup | 10K 10K 9K
7K 10K 9K 6K 13Gup | 9K 9K 8K
6K 9K 8K 5K 12Gup | 8K 8K 7K
5K 8K 7K 4K 11Gup | 7K 7K 6K
4K 7K 6K 3K 10Gup | 6K 6K 5K
3K 6K 5K 2K 9Gup | 5K 5K 4K
2K 5K 4K 1K 8Gup | 4K 4K 3K
1K 4K 3K *1D 7Gup | 3K 3K 2K
*1D 3K 2K 2D 6Gup | 2K 2K 1K
2D 2K 1K 3D 5Gup | 1K 1K *1D
3D 1K *1D 4D 4Gup | *1D *1D 2D
4D *1D 2D 4-5D 3Gup | 2D 2D 3D
5D 2D 3D 5-6D 2Gup | 3D 3D 4D
6D 3D 4D 6-7D 1Gup | 4D 4D 5D
7D 5D | 6D
8D 6D | 7D
9D 7D | 8D
10D | 9D

This table is extracted from Sensei's Library: World Wide Comparison.

So, if KGS has to recovers this 4 stones at the 9k level... why didn't they leave the old system? Sincerely, I don't know the answer. I don't mean that KGS2 rating is better than KGS3, but it may have been more "real".

Talking about "better"... I think that none of the KGS rating systems are good. For example: how can you improve your level at KGS? By NOT playing.

In fact, my original account at KGS "alejo" showed this image today:

Which is something obviously wrong. I haven't played with this name for a couple of months and there are only 3 games listed for the last 6 months... As you can see, inactivity is the best way to increase your rating.

The real fact is that I'm at 9k. Yes, 9k at KGS, which means that I don't know where I'm at all.

Trackbacks

    No Trackbacks

Comments

Display comments as (Linear | Threaded)

    First off, I don't really see the need to have some sort of ranking system that works the same everywhere. In fact - I think that this would be impossible. Isn't it enough that rank guides us towards finding an interesting game with a suitable oponent?

    Looking, for instance, at the EGR (European Go Rating) system there is ample evidence that this system suffers from geographical variance; I've heard rumors about similar circumstances in AGA. It is possible that there might exist such discrepancys on KGS too. E.g. people only playing certain hours of the day, mostly playing people in certain rooms. But this is only speculation. The point is that these disturbances occur and they are not easy to manage.

    That said.. KGS ranking system certainly is weird. Some players around my level are unbeliveably weak/strong. This seems very strange to me.
    #1 pel on 2007-05-28 10:58 (Reply)
    Weird... very weird indeed... X-files... hehehe...

    "Isn't it enough that rank guides us towards finding an interesting game with a suitable oponent?"

    Yes, should be, but you also recognise that you find people who are incredibly strong or weak at your same level, which is something that the rating system should avoid.

    Now, I will appear as a weak 2d player. Toooooo weak, indeed.

    This is one of the consecuences of this bug of KGS... what should I do? Losing against bots till I reach something near 9k? Lose against a 1k player who will start thinking that he has beaten a 2d, affect his rating and so on?
    #1.1 Alejo on 2007-05-28 12:39 (Reply)
    There is no such thing as an absolute rank, it's all relative within one ranking system. For example in Korea, there used to be no amateur dan ranks and a 1 gup (=1 kyu) was the strongest rank an amateur could get. The only real measure of a Go ranking system, in my opinion, is whether the differences in ranks are an accurate prediction of the proper handicap needed for a good game. For this purpose, the KGS3 ranks are much much better than the KGS2 ranks were. In fact, around my level (KGS3 1k) many people stopped playing handicap games altogether with KGS2 because white was at such a disadvantage.

    Your criticism of how ranks tend to solidify when you play a lot and become crazy is valid. I don't see the problem of getting weird ranks when you stop playing however - after all, you are not playing, so why care about that rank? (it even gets marked with a questionmark for everyone to see that that rank may be wildly inaccurate) - but for players who play a lot it can take a long time of winning the vast majority of games to move up sometimes.
    #2 nik on 2007-05-28 11:29 (Reply)
    Of course there isn't an absolute rank (though it'll be nice) and we just play for fun. But it is logical to think that an European 10k will beat a 13k without handicap... while both are a 9k on KGS.

    I understand the problem with KGS2 and handicapped games... hadn't thought of it... but, anyway, isn't there a better way to fix this problem at the 9k level? For example, following an average between the other ranks, the "weakest 9k" should be around 12k; just increasing one stone per level... as usual, as all the other ratings do.

    And I'm sorry, but I don't accept that my rating increases even if I don't play, it should be somewhat stable. I understand the method they use for calculating your rating based on others', but this leads to such incredible situations:
    I beat someone who is 15k... then I'm, at least, 14k... Next game I lose against a 13k, then I'm 14k... Two months later, the 15k player has reached 3k level and this affects me in a way that I become somewhat like 7k.

    I still dislike those two problems of KGS ratings...


    (Anyway, I'm going to modify the size of the "commentary writing", they are too small...)
    #2.1 Alejo on 2007-05-28 12:33 (Reply)
    Well, finally solved the problem of the comment's size.
    #2.1.1 Alejo on 2007-05-28 13:08 (Reply)
    I like the new KGS3 ranking system, as it seems quite close to real life AGA rankings.

    I think the problems that made you a 2 dan could have happened on KGS2 system.
    #3 Nate on 2007-06-18 08:09 (Reply)
    Yes, I agree that this could have happened on KGS2. In fact, it already happened to me at the very beginning... but I expected it to be solved on the new ranking system, partialy, at least.

    Maybe I was expecting too much...hehe
    #3.1 Alejo (Homepage) on 2007-06-18 08:15 (Reply)
    There is no such thing as KGS2 and KGS3 ratings. It's a myth. There had been similar adjustments before and there are going to be others when needed. There had been a slow drift towards the bottom of the scale and the adjustment that occurred at the same time KGS3 was released just restored the previous ranks. It's been almost a year why are people still bothered by that?

    As for the validity of KGS ranking algorithm it's very good but only if you play a lot, if you don't play enough the system has not enough data and in the kyu ranks that creates some overrated players.
    #4 Codexus on 2007-06-18 14:55 (Reply)
    The own KGS server calls it KGS3... Maybe it's just an adjustment (I don't really know), but in order to identify this "new" period the go comunity calls it KGS3.

    It's been a long time since they modified it, I know. But the point is not arguing about KGS2 and KGS3, the real endpoint of this post was to criticise how KGS algorithm deals with players who don't play much.

    And, in fact, if there had been a "slow drift towards the bottom of the scale" I think it could have been better... Currently KGS3 overrates players compared to all other real-world ratings on the lower levels.
    #4.1 Alejo on 2007-06-19 03:33 (Reply)

Add Comment


Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
Standard emoticons like :-) and ;-) are converted to images.
E-Mail addresses will not be displayed and will only be used for E-Mail notifications