Table of contents |
nkpuxa Yes, to add my experience, 3d on IGS is much stronger than 1d KGS, at least. KGS also has many 9d that are nowhere near as strong as 9d on other servers.
seven I agree that there is something wrong with the KGS-ranks. I have been playing both IGS and KGS. On IGS i play around 17k+, on KGS around 17k. I have added just that 17k-rank for KGS but there is no balance for the KGS-ratings now. Are there any other higher ranked players playing both IGS and KGS which can add more points on KGS, then we can extrapolate
Calvin Well, I am IGS 8k and KGS 15k as of December 3, 2005. I think there is a large standard deviation in strengths among double-digit kyus on KGS, just like I noticed on IGS when I was weaker than IGS 20k*. I've also noticed a relatively large number of new accounts on KGS, which will naturally throw some chaos into the system.
Stormer igs 6k+ kgs 8k.
I wonder if the recent addition of rated games for bots has destabilized ranks surrounding it. Most recent versions of gnugo end up rated 12-14kyu on kgs. Rated games can have 6 handicap stones. In the kyu ranks, high handi games could be skewed in favor of those recieving the handicap. Many players on both servers are reluctant to give a large handicap in rated games, but gnugo will give all of the handicap you want. It could be that people who play the bots a lot could end up with a higher rating than other players who tend to only play against humans. This could lead to some more variance in the ratings.
One piece of data that seems to support this theory is that the only gnugo bots to maintain 12kyu on kgs is set up so as to play only even games.
How old is the above? As of 9 Jan 2006 there are several 12k bots on KGS which are happy to play ranked handicap games. There is also a blitz bot (Wahabot) which rose from 12k to 10k after moving to 25 moves/minute play, though I can beat it without much handicap. --Damien
In any case, kyu ranks are always going to be a bit wacky. And I think the "ranking bots" do more good than harm. Its a pain for new accounts to get people to play them rated games. So it seems like a nice service. Also useful for the people developing go bots.
On KGS I'm a 16k; in my college club I've had close games with a AGA 1k at 9 stone handicaps though. A few years ago I seemed to be about 12k in the San Francisco Go Club, but 16k on IGS, which matched what the club members told me to expect. I also know a 9k KGS who's been playing around 2-3k in tournaments, -- Damien
I can confirm that. I know at least 5 people playing 4k in German Tournaments, who are 10k on KGS.
Reflame I think there is an easy way to fix the differences among ratins systems - why do associations not do it this way?
Jan, 2006: Recently entered a tournament at my local go club. I was not an AGA member yet so I had asked some advice. Some said I, as a 12K on KGS, should enter as an 8k, others said 10k. I entered at 9k. I won three games by over 30 points (giving and receiving handicap) and lost one (getting 7H) by about 6 points. The table above suggests about 6k for AGA, but that seems a bit excessive. 8 or maybe 7 might have been more reasonable. Additionally, when I was 15k KGS I gave a 15K AGA four stones for what felt to me like an even game (we played three games). --erislover
I run a club on the east coast and from the members of my club it goes something like this: 6d AGA = 8d? KGS, 1d KGS = 5d AGA, 3k KGS = 2d AGA, 8k KGS = 4k AGA, 12k = 8k AGA, 16k KGS = 16k AGA. IMHO there is only a good correlation between ranks from 1d-12k. Anything above 1d KGS has an inconsistent correlation and anything below 12k is inconsistent as well.
The main reason for the problem is that at 12k people can fight bots, which frankly have serious holes in their playing. 12k is therefore a breaking point as far as rank. Above 1d has two problems. Mainly that 1 stone per dan is a fairly steep handicap and that there are not as many high dans playing. Regardless, you never know when you are playing someone who is watching TV and in general I think its a good idea to not focus on rank too much.
I stopped playing on IGS a couple years back, (5k) at the time, mostly because, at the time, handicap games were rare, and it seemed as though I only got to play against other 5k's. I moved over to KGS and continued my progression there, up to 3d and then the rating change hit, and everyone dropped 2 ranks. So Now I'm 1d KGS.
A friend of mine started playing on IGS recently, so I figured I'd give it another shot. Turns out I'm between 3-4d on there, which does seem to hold with the above chart. Perhaps this has to do with the fact that IGS, I believe, has a predominantly Japanese constituency, it may be more reflective of their system of ratings. Whereas, while KGS may be more of an American Phenom. I highly doubt that many KGS players have actual AGA ranks.
According to the
KGS World Map, KGS is mostly a European thing, and European ranks are considerably stronger than AGA, accordign to the chart above.
evpsych: What's the +/- on this? I was under the impression that a 10k nngs is about 17k igs or thereabouts. This table says 10k=10k. +/- 4 stones?
Alex Weldon: I don't believe the ranks given for Gup are accurate; Koreans tell me that nowadays, you go 1 through 7 dan after passing 1 Gup. Perhaps it's been adjusted to match the kyu system
Most online Korean Go servers rankings are inflated by as high as 10~5rankings. However if you play at a Korean club or academy the ranking with an instructor who gives out the rankings it is almost similiar as the table
(July 24 2003) Isn't an update of this list needed now that NNGS has raised rankings by 2 stones? (see help new-ratings on NNGS).
StormCrow: Persuant to the above comment and the bumps in the ranking graphs on NNGS, I've updated the table by increasing all the NNGS rankings by 2 stones.
Alex Weldon: I think either the rankings for KGS, or those for IGS, as given above, are wrong. I was 5k* IGS many months ago, then stopped playing and starting playing on KGS. I lost my first few games, and my rating actually ended up in the low double-digits. I was underrated for sure, and quickly improved back up, but progress has slowed down and I'm still only 6k on KGS (still slightly underrated, since I win about 70% of my games). Meanwhile, my actual skill has improved, so when I decided to try playing on IGS again today, I demolished all my 5k* opponents. So, although the table above shows that KGS ratings are inflated relative to IGS, I think one or the other server must have changed their system, because it seems to be the other way around in my experience.
On the other hand, it may just be a difference in style... maybe my style works better against IGS players. Who knows?
UrbanNilsson: According to the KGS club Tenuki at
http://www.abstractboardgames.com/kgs/add.php, the ranking difference between EGF and KGS is the other way around. They say that 20k KGS is equal to 15k in Sweden (and Swedish ranking is within a third of a stone of EGF's according to
http://www.european-go.org/rating/rgdrep.html)
HolIgor: This may be true about the double digit kyu range. Novices on the servers honestly start at 30k and then painfully climb up the ladders fighting with the similar players. Their rating usually lags behind their real strength. In real life promotions happen very fast depending on the results of the tournaments or when stronger players feel that they have to decrease the handicap in order to have a chance of winning. Yet at the level of single digit kyus and dans the table is in my opinion quite correct.
May I remind you that, AGA and EGF rankings are based on wins and losses. Korean, Japanese, and Chinese rankings are based on skills and concepts learned.
I am KGS 2k and AGA 1d, I also just played (apparently) a Korean 16 kyu (or Gup I'm not sure, he said kyu) who was clearly about two stones stronger... I was floored, are koreans THAT much stronger? ---AdHoc?
I believe that the table may have to be changed; recently DGS has been changed so it is impossible to go under 30 kyu. (Before, the minimum was 40 kyu.) Has this been adjusted? Or have I been asleep? ---DanielB
As far as I know, official EGF ratings start with 20k, not 28k. --gimpf
ilan: The new KGS ranks are really harsh. I believe that a 5p player on KGS is now only 4p.
Chris Hayashida: How did that happen? Pro ranks are outside of the rating system...
Velobici: ilan meant to say The new KGS ranks are really harsh. I believe that a 5p player on KGS is now only 4p. ;)
Velobici: I have attempted to update the KGS ratings based upon several people I know (AGA 3D - 10 k) and the information posted at http://westerngo.org/ladder/
Does anyone have a good idea about how Dash ranks fit in with the others listed?
Malweth: Perhaps Dashn and LittleGolem should be added? Yahoo perhaps should be removed from the list since ratings there are unstable. I'm not sure how these should be compared - especially since my AGA and KGS ranks are somewhat unstable.
ilan: I played exclusively on Dashn for 3 months and I can report that the ranks should be considered separately from the other servers. The reason is that the majority of players at all levels seem to be somewhat ignorant of opening principles and engage in unreasonable early fights. Therefore, if you know how to deal with this, you rank will increase almost without bound, but if you don't, your rank will be dropping fast. So, I optimized countering this strategy with a very solid style and I went from 8K to 1D in 3 months. Likewise, a KGS 1 dan told me that he is 5 dan on Dashn. Now, I'm starting to play at Cyberoro and the majority of players there do not play these unsound early attacks, so my slow opening is putting me at a disadvantage, and I'm having trouble keeping at 1K and I am now (very temporarily, I hope) 3K. So, Cyberoro must be about 2 or3 ranks stronger than Dashn, just because of the greater diversity of styles you must learn to deal with. According to minue622, a KGS 5 dan from Korea, the current rankings on KGS are fairly close to the ones on Cyberoro. The conclusion is that Dashn ranks should be about 2 or 3 ranks weaker than the current KGS ranks.
[102] (Sebastian:) If this is really true, what about the correspondences in between 4D and 7D European - to which Japanese ranks do they correspond? Would it be possible to extend the table so that professional ranks are included? I heard, on IGS there is quite an overlap, and I seem to remember that 1p=5d there.
Fwiffo: With the recent ranking system changes at KGS, this chart is no longer accurate. One barometer for comparison would be GNU Go based robots. Most were about 9k (close to 8k) before recent changes, now most are 11k. Allegedly GNU Go is about 12k on IGS. At the higher end fo the spectrum, most dans seem to have lost about a stone (e.g. TheCaptain dropped from 5d to 4d).
Malweth: It seems as though Velobici changed the KGS ranks recently. Are these ranks accurate?? If so, I really don't feel bad about my 7-8k ranking anymore! LOL - Time to play more rated games on IGS! (I still don't quite believe this - maybe I can test it out a bit).
Velobici: Yes. I did change the KGS ratings based upon those that I see in the Western Go Ladder ladder and both mine and my son's ratings. My son plays on both IGS and KGS so he is a comparison point for me to use. At his level, AGA 3D, KGS is harsher than IGS. At my level, KGS has become brutal in its ratings :) Please do test this out...either confirmation or refutation is greatly desireable. Oh! for a world-wide pro/am numeric rating system that would let me see that I can beat Yi Chang Ho with only 15^H^H25 stones :)
Stormer I'm a kgs 10kyu. It seems that ranks around mine ended up shifting to be about 2 stones stronger. I can't imagine this can account for the differences in rank shown here. Just the other day I was talking to an IGS 2k* who plays occasionally on KGS, he said that IGS ranks are still about 2 stronger than KGS. Of course he does not play enough on KGS to have a solid rating, so maybe take that with a grain of salt.
DrStraw I think these changes to KGS are totally wrong. As a long time AGA 5D I have very little trouble maintaining 4d on KGS if I play seriously (with occasional forays into the realm of 5D) and can easily maintain 3D on accounts I use for playing just for fun. In my opinion, at least at the dan level, KGS is less than 2 stones stronger than AGA. I would think it is comparable to IGS now, with IGS being a little stronger. I rarely play on IGS but the account I do have is 3D.
Velobici: Steve, could have improved? Your last AGA rated game was over 3 years ago. Your sigma of 0.36 will tend to pin your rating in place, coercing the rating of others to change in place of yours changing. This combined with standard AGA slaughter pairings may make your rating a significant underestimate. But please play lots of games on KGS and IGS and AGA tournaments and change the table to match! Or just change the table. ;)
Perhaps a look at tiers 5 and 6 of The Western Go Ladder
http://westerngo.org/ladder/ would shed some light on this.
DrStraw: I think the root of the problem is that AGA ratings are based on tournament games, where everyone is trying their best. My experience on KGS tells me that only players at 6d and above consistently play this style of game (there are a few expections) and that ranks of 5d and below are based on a mixture of fast games, fun games and serious games. This would tend to make the results less predictable. It would also have the effect of "strengthening" the KGS ratings because an AGA 4d would tend to be consistent relative to other AGA ranks players of comparable or lower rank, but would tend to lose more games to stronger players, thus forcing his KGS rank down. At lower levels this effect is probably even more pronounced as the majority of games played by kyu players could hardly be considered serious in the sense that AGA rated games are.
isshoni: i can only speak of KGS vs Japanese S/DDK: on KGS i 'm about 13-14k. According to table above, that would be around Jp 5k. In Japan, i don t have any official rating, but occasionnaly i go in salons and always get to play the 'sensei' of the place. They recurrently assess me between 3k and 1d, even though i must say that their judgment is based on very few games. The table would rate me 9-11k on KGS.
ilan: You can see my KGS rank graph here http://www.gokgs.com/graphPage.jsp?user=ilanpi
I have played very few games there since August, 2004, so there is a sense in which your actual KGS ranking, according to the pre September algorithm, can be computed by using the inverse map of the function defined by my post August KGS rank graph.
eng60340 is the statement kgs 12k == igs 7k accurate? the last time i went to igs (was a long long time ago) i had the impression that igs ranking were stricter. i.e. kgs 12k == igs 16k etc.
Malweth: Well... Perhaps they are ok. I just beat an IGS 1d as white (W+T, but the game looks fairly even and it wasn't blitz). I was 7k on KGS most recently, but I haven't played KGS in a while (except for teaching games).
Velobici: another tidbit regarding KGS ranks. Argentina's representative to the 26th World Amateur Go Championship plays on KGS as quaker. His rank on KGS is 3 dan as of June 2005, down from 5 dan in October 2004.
Malweth: The Interpolation of KGS Ranks based on BGA ranks (August 1st, 2005) fits for me: 3k AGA (7/21/05), 8k KGS.
(moved from rank) CBlue: The beginner ranks are called 'class', while master ranks are called 'degree'. In Japanese, those are Kyu and Dan. In Korean, it's Gup and Dan, and in Chinese it's Ji and Duan. The internationally most commonly used terms are the japanese expressions. 'Kyu/Dan' ranking system is also used in martial arts such as Karate and Judo (black belts indicate dan ranks, while coloured belts indicate kyu ranks).
Kyu ranks usually start at 20 (beginner) or in systems of finer scale at 30, Gup and Ji ranks start at 18. The class ranks go up to 1, which is the strongest class rank. If the player further improves, he enters degrees and starts at 1d. Amateur degrees go up to 6d or 7d.
Professional players are measured in so-called Pro-Dan ranks, which are indicated by a 'p' instead of 'd'. Those go from 1p to 9p and are not as directly related to playing skill as the usual ranks, but players can get promoted for simply playing a large amount of games or winning a certain sum of prize money, depending on the country's Go Association. In order to become a professional, players have to take a very hard exam.
On KGS, beginners start at 30k (kyu), and amateur dan ranks are not limited to 7d but go up to 9d which allows a finer scale.
It is said that 1p corresponds to 6~7d amateur strength, but especially in Korea the competition between Inseis has become so hot that freshly graduated 1p players there are probably a lot stronger even.
To complete the translations, Go is actually called Igo in Japan, Weiqi in China and Badook (also written Baduk) in Korea.
Niklaus: Isn't the "unit" of the ranking system supposed to be handicap stones? So the only difference between the different scales should be where you set the bar for 30k (or 29k, since everybody below is 30k), and that difference should be the same on every level onwards? Most of the chart above is more or less consistent, but there are some areas (such as IGS double digit kyus and Japanese high dans), where the "unit" seems to be more or less than one handicap stone.
amc: I have a question about ranking. I know that "theoretically" a difference in one rank is equal to one stone, right? But what does this mean exactly? Does it mean that two players of different ranks playing lots of games with the appropriate amount of stones should split the results 50/50? But if this is so, isn't this more or less easy to determine? What exactly causes the differences between the different ranks? (servers, regional association, etc.)
Answer (Charles Matthews): Yes, it is assumed that for purposes of playing handicap games the correct rank difference is the number of stones to give 50-50 results, while one stone too many or too few will give more like 67-33. There is some bias for small numbers of stones in that a two-stone handicap is more like 1.5 stones in practice. The observed difficulties in cross-calibration of ranking systems will stem from a number of factors, such as: different pools of players worldwide not strongly interacting; lack on some servers of handicap games played; styles and depth of thought being different for quick server games and tournament play.
W: What about differences in style? I do not think a linear rating system will ever be completely accurate. I am sure there are cases of players A, B, and C, where A can give B two stones and play even, B can give C two stones, and C can give A two stones, if the styles of play and focus of learning are different enough. How would you rate such players easily? Or do these situations not occur at higher ranks?
anonymous-- At best ratings measure some sort of average performance over all go players in some large group so there is no certainty and no contradiction about how two individuals would play based on their ratings.
Could anybody please explain what GoR is? It is used in http://www.european-go.org/rating/rgdrep.html. Thx! -- Sebastian 2003-09-10
Andrew Grant: GoR refers to the European Go Rating system, which is similar to the Elo ratings used in chess. The results of every European go tournament are sent to the organisers of this system (in the Czech Republic), who enter them into their computer which analyses the results and produces a ranking list every month giving the current rating of every player in Europe who has played in a tournament recently. Traditional kyu / dan grades are meant to be 100 points apart on this system: 2050 is an average shodan, 1950 is an average 1 kyu, 1850 equates to 2 kyu, and so on. You can access this list from the EGF website.
Hans: One small correction. An average first dan has 2100 points, an average 1 kyu 2000, etc. At least this is stated in the introduction to the system on the site:
http://www.european-go.org/rating/gor.html
Andrew Grant: Sorry, but this is wrong. It was true when the GoR system was set up but grades have inflated since then. This month (September 2003) the average GoR of European players claiming to be shodan is 2048.
tderz: (several years after Hans' and Andrew?'s entry:
- 1) Andrew, where is your source ? (Hans gave one)
- do you 'mean' average = arithmetic average, mean or geometric mean or median or (...) ?
- 2) perhaps you wanted to criticise Hans for his statement 'average first dan has 2100 points,(...) ' where in fact he quoted it precisely from
An average 1 dan should have Go Rating about 2100. Reading this statement one doesn't know whether this is a desideratum, a description of the status quo or a rule, where the player has to conform with.
This rather means to say something like: 'a Go shodan is set at 2100 ELO points'.
This could serve as an anchor. (I do not know which players serve as an anchor in GoR)
- 3) so that you state that the (arithmetic) average, (mean) shodan has/had 2048 ELO points at some time is due to inflation, is some reason, but might not be the only and predominant one.
First of all, it is a momentarily picture of that rank population and their ELO points, perhaps differing from the next months for the same population (most probably not by 52 points, though).
Secondly inflation is not a single reason (it is expressed here in one number '2048'), it is a syndrome, rather more probably it has several reasons. Actually other reasons could be there as well, e.g. 1 dans promoted away to 2 dan, leaving unpromoted & unimproved shodans behind, yet with an influx of promoted 1-kyus. (in total that's also an inflation theme).
This introduces a chicken-egg question: what was there first the ranks of players, enrolling in tournaments, producing ELO ratings, which then indicate ranks. Which rank to take as guidance? Again it will be a mix of several. Once should become suspicious if the sytem were to produce 9+ - dan amateurs or move in that direction. I guess/hope that the system uses the top amateurs as anchor for calibration, say as 7-dans (2700 ELO points). Then the whole system should not be so much off (Jie Li 9d et. al create problems, though), but as least stable, under the assumption that the the top amateur dans play consistently aslo during their whole career. It was nicer if there was a single rating list comprising professionals and amateurs.
gimpf: That's something I didn't understand when reading the EGF docs... I thought, the dan-kyu grades are given in respect to the rating - but when looking at the ranking-list of all the players, grades seem to be given a little bit at random :) Why is the rank not steadily changed according to the rating?
Stefan: The ranking list doesn't give grades. It merely lists what grade the player used in the last tournament. Taking my own name as an example: I am "officially" a 2k in Belgium, I registered in the last tournament I played as a 1k (for stiffer competition) and my points are what they are.
Andrew Grant: In answer to gimpf, kyu / dan grades don't necessarily reflect GoR ratings because the EGF has no power to compel players to play at any particular grade. Nor should they have this power. People should be free to play at whatever grade they wish.
And to follow up my answer to Hans, ranking inflation is inevitable in any system such as the GoR rankings, unless steps are taken to counteract it. The problem is that at any one time there will be a significant number of fast-improving players, but few players who are declining in strength, and those will only be declining slowly. Most players' strengths are static. So if a few players' GoR ratings are increasing rapidly that is paid for by a slow decrease in the GoR ratings of the established players. Of course, the established players aren't really getting weaker. They're just being beaten en route by keen young players on their way to greater things. But the effect is to lower their GoR ratings just the same.
Hans: I understand that the rankings are slowly inflating because of fast improving players. Nonetheless a player who enters the rating list as a first kyu is given an initial rating of 2000. Besides they have built in a small factor, called e, in order to prevent the rank inflation, resp. rate deflation (depending on the way you look at it) (The details can be found in the description where the calculation of the new ratings is explained).
I use the following formula for those below 4kyu Real Life = 4 + (0.75*(KGS -4)) this seems vaguely accurate
hotcoffee: What was the source of the data for the rank comparison table? Can more information be found for KGS rank comparisons?
excession: You might be intersted in this: http://kgs.goclub.info - its a BGA / KGS comparison
Intetsu I think the rank comparison is way off. Im 2 dan on KGS while only 2 kyu in the AGA. According to the chart I should be 6 dan, I wish.
Velobici: Could your AGA rating be off? Please compare with Landon Brownell 6.9 AGA. He is LBgo on KGS 2d.
nelsson I gathered some ranks said by people into one clump. I also deleted their messages to make this page clearer. Kill me if I did wrong.
KGS 2d, AGA 2k (?)
KGS 2k, AGA 1d
KGS 2k, EGF 1d (
KGS 3k, EGF 1d
KGS 3k, Germany 1k
KGS 6k, AGA 1k
KGS 6k, AGA 2k
KGS 6k, AGA 2k
KGS 6k, EGF 3k
KGS 6k, EGF 4k
KGS 8k, AGA 3k
KGS 8k, AGA 11k (rank for one tournament)
KGS 9k, AGA 7k (First AGA tournament)
KGS 9k, EGS 6k (EGF?)
KGS 10k, IGS 6k
KGS 10k, Finland 6k, Denmark 7k, 7K EGF (rank suggested by GoR)
KGS 11k, AGA 14k
KGS 15k, IGS 18k, dashen 18k
Table moved to Rank Worldwide Comparison/Data Collection instead.
Amlodhi: I think one thing people are also forgetting about the difficulty of accurately measuring AGA or EGF against KGS or IGS rankings is that a lot of people play much better or worse online. Most of the younger players I know play better online, and it works in the opposite direction for most older players, who learned how to play strictly on a goban.
Speaking for myself, my AGA and KGS ranks are both 9k, and my IGS rank is 8k, which I attribute mostly to a more aggressive style by the players on IGS, where the focus seems to be more on tactics than strategy, and I usually play well against that kind of all or nothing style of play.
Malweth: wouldn't this be useful as a database that people can update as things change? I can set it up w/o too much effort if people are interested... (might make a nice first Ruby project for me) RobFerguson: Sure go for it, post a link. The results might be interesting.
kevinwm: I haven't seen any comments on how the KGS rating anomalies might be related to the fact that KGS' ranking calculations are different from the way AGA and other associations calculate rank. Since KGS assumes that the probability of beating someone 1 rank greater is mostly constant over a large range or ranks, maybe what we're seeing is that this assumption is false! Based on the tight grouping of KGS players around 13-16k, corresponding to a wide range of AGA rankings, it would seem that for the corresponding AGA rankings, these players would have a greater chance of beating someone who is 1 stone better (AGA) than them, compared with a 1d or 20k AGA player.
In other words, in order to have a 67% chance of beating an 8k-12k AGA player, you'd have to be significantly higher rank, while this does not appear to hold true for 1d or 20k players.
BobSmith: The Korean rankings are way off. I am currently in Korea right now, and I play at a local Kiwon (Go Club). I'm a 7 gup there, and I am a 13 kyu on KGS.
amadis: Here are current ranks of some top European players:
Euro-
KGS
pean
v.3
Alexander Dinerstein (backpast) 8d 10d Cho Seok-Bin (bin7674) 8d 11d Cornel Burzo (Cornel) 6d 8d Franz-Josef Dickhut (fj) 6d 8d Pal Balogh (macisajt) 6d 8d
Looks like the new KGS ranks are still a bit off from the European ones.
Calvin: This table is out of date. The KGS3 ranks seem to cap at 9d as before. And I don't think any European amateurs claim higher than 7d, so where do these 8ds come from?
amadis: These ranks are current as of September 2006.
KGS ranks nominally top out at 9d, but if you look at the
graphs you will see that the some of the ranks are actually higher than 9d as computed by the system.
I assigned EGF ranks according to the players' current ratings. According to
Aleš Cieplý, who devised the system, "An average 1 dan should have Go Rating about 2100 and the difference between grades is set to 100 (6 dan = 2600, 1 kyu = 2000, 20 kyu = 100 etc.)."
Dinerstein is a pro, so of course he might well rank higher than top amateurs.
I asked glue whether the new KGS ranks are meant to correspond to any real life system. She said no.
On the other hand, the room message in the Deutsche Ecke on the beta server says, "Bitte beachten: das Rangsystem von CGoban3 wurde verändert, um es an EGF Maßstäbe anzupassen." That is, "Please note: the ranking system of CGoban3 has been changed in order to align it with the EGF scale."
What gives?
kevinwm: I think the explanation about why KGS ratings are off makes a lot of sense. In the mid-k range, there are a lot of large fights that players are just learning to fight. A big win or loss in one of these fights can make a huge difference in points - so that 1 stone doesn't mean as much at this level.
Velobici: A single fight can result in a large point difference in a single game, but the rating system is based upon the number of games won or lost, without regard to the point difference in the games.
A rating system adjusts ratings in light of game results to increase the probability that the rating difference will accurately reflect the probilility that the stronger player will win a particular game. When the stronger player wins a game in which he is a heavy favorite (relatively large rating difference), the ratings of the two players change little. By contrast, when the stronger player loses a game in which he is a heavy favorite (relatively large rating difference), the ratings of the two players change considerably.
In other words, a rating system is a post-facto attempt to mathematically deduce relative strengths (expressed as rating differences) from a set of win/loss results.
amadis: (28 Dec. 2006) I added comparisons of high dan KGS ratings vs. AGA and EGF. Here are the data:
KGS AGA EGF bin7674 9 8 roln111 8 7 HYang 8 9 bigbadwolf 8 8 darkmage 8 8 macisajt 8 7 Cornel 7 8 6 Cpop 7 7 toxxicu 7 7 jim 7 5 Lacour 7 5 SanKingTim 6 5
Admittedly, my data are rather thin. If you have more, please provide them.
Then I smoothed out the lower dan and single digit kyu values for KGS, based on the observation that KGS kyu ranks have inflated by a couple of stones since September.
I left the lower kyu values alone. I have no idea whether they are really as compressed as they are shown in the table.
Also, I left AGA ratings as EGF+2. They may be closer to EGF+1.
quantumf 13 March 2007: Any cyberoro updates? I've played about 100 games on cyberoro, and have stabilized around 17k (I entered as an 11k but kept losing until I reached my current level). On KGS, I'm around 4k/5k. This is an astounding difference!
Velobici: CyberOro appears to have a very different playing style than KGS. Until your strength in these areas improves, your CyberOro rank/rating will continue to lag your KGS rank/rating. But just think how strong you will be will your the two converge!
209.51.253.140: Does anyone has experience of how the OGS ranks are in comparision with the others mentioned here? And/or could add them to the list?
AhSayuni About CyberOro. I've been playing there for quite some time though not very often (I've had around 60 games I think) I signed as 5k, which was my rank on the old kgs, and had a pretty easy time getting to my current rank of 3d which is, at least, solid. On the new kgs though, I am no stronger than 2d, quite a solid 1d actually. I've heard that generally the low ranks there are relatively strong to other servers.
[mdm:] 1 August 2007: I'm 2k on KGS, but a "solid" 17k on Oro.
tapir: Hi, I used to play as 18k/19+k on IGS, then made a new account on KGS. About a month later (reading Kageyama and Attack+Defense in the meantime) I started playing some more on KGS. Instantly rising up to 5k (later lost this rank but regained it by now). (The ranking table gives 11k as equivalent of 18k IGS)
I felt with only a few low level players on IGS, playing each other all the time there is no real measure for improvement. All players improving somehow, those who improve faster may rise a bit those who improve but not as fast as others may remain at the same rank or even lose it.
Bill: This kind of inefficiency can affect any server where nearly all games are even games, not just IGS. In particular, groups of new players can play against each other and improve without any of that improvement being reflected in their ratings. In the worst case, they may have trouble getting handicap games from stronger players who suspect that the new players' ratings are too low and do not want to lower their own ratings.
tapir: However I felt this problem is more grave on IGS because there are less double digit kyus there and no bots like on KGS, which don't improve at all. (Only some players with thousands of games, but still ranked 20k... but they are not bots as far as i know.)
PeterHB: DDKs have apparently been upranked by 2 or 3 stones on 2007/03/27.
pomegranat?: It appears that IGS did away with all ranks weaker than 17k. Am I alone in thinking this is a bad idea? It just seems that it would discourage a lot of weaker players from really getting into Go (on IGS at least), since it would be next to impossible for someone who's just starting off to find people around their strength and adequately chart their progress. While 17k isn't incredibly strong, it still does take a good deal of effort and knowledge of the game to get up to that point -- something I think that a lack of equal opponents would make all the more difficult.
tapir: Now there are less beginner ranks, so a greater part of the player population is around your rank. A larger population reduces the problem of deflation mentioned above - a population of players playing each other, improving, but staying at the same rank all the time. (Taking for granted the sad reluctance to give correct handicaps.)
And those who are now "17k" are still the same players being "22k" some days ago, they are in no way stronger opponents (maybe more self-confident now :)) than before. And I don't see how this lessens the number of equal players at each level.
pomegranat?: I wonder if one possible way to solve this would be to have an 18k-30k beginner ranking pool that is outside of the standard ranking system used for everyone stronger than 18k. That way, beginners could find people around their rank that are more accurately ranked than just "," but still avoid causing ranking drift for everyone else. After hitting 17k in their own rating pool, or beating a certain number of people in the main ranking system (which I guess would have to be free games for the people in the "main" rank pool), they'd then move into that pool of ranks. I think my main concern here is that 17k has become a huge pool of players who all can't really play even matches -- 'I remember playing 21k's when I was a 17k, for instance, and noticing a fairly large difference in skill -- such a difference in skill that are now lumped together.'
tapir: No, they are not. Those 22k/21k are 17k now, the 20/19k folks 16k, those 19+/18/18+k are now 15k (that happened to my account) etc. etc. It's definitely not lumping together all those former being between 17k and 22k in one rank. Only the rank difference is less now. 22k / 19k was three then, 17k / 15k ist two handicap now... It might even encourage the players playing more handicap games.
Harry Fearnley: I have met several Koreans who know what their (Amateur) Dan ratings are. Recently, I met some players who claimed to have no idea what their "Gup" ratings were.
Is there a source of reliable information on the correspondence between Korean (amateur) Dan ratings, and other worldwide rankings?
How do those Koreans who use Amateur Dan ratings describe the ratings of weaker players? ... as "Gup"?! If so, ...
What is the status of the "Gup" comparisons shown on the page at Rank Worldwide Comparison?
Dieter: see gup.
Anon Linguist: I don't know if this is just random silliness, but I've seen two 23kDGS get rated 15-16k by it, a complete newbie 1D, and a not entirely complete newbie 13k. It seems slightly unreliable, or I'm lucky to be in the company of several extremely talented beginners. The 1D is a bit of stretch, though. Perhaps it could be inferred that less newbieness means a lower score, but that rather flies in the face of what may be the intent of that test.
EGF (age) | KGS | IGS | Dinerchtein test ----------------- | ----- | ----- | ----------- 5 kyu (3 months) | 2 kyu | n/a | 4 dan 11 kyu (3 months) | 7 kyu | n/a | 3 kyu n/a | 1 kyu | n/a | 4 dan n/a | n/a | 5 kyu | 3 dan n/a | n/a | 11kyu | 1 dan?! 1 kyu (1 month) | 1 dan | n/a | 4 kyu n/a | 6 kyu | ... | 7 kyu n/a | 4 kyu | ... | 4 kyu n/a | 4 kyu | ... | 11 kyu
An ego booster. :) (Nicks/names left out to discourage any bragging-factor. "Age" is time since last tournament.)
unkx80: I tried it too. Was told that I am European 5 dan. Like, wow, so overrated. The biggest problem is that it is MCQ. If it is not MCQ I might never thought of some of the choices.
Anyway, I hope that he releases his answers (together with the reasoning) later so that it can be valuable study material.
Phelan: I've already posted this at the CoffeeMachine, but thought this was an appropriate place as well: I saw a post on rec.games.go which I assume was by breakfast, saying that he will explain the test problems in http://play.baduk.org/, this sunday, Jan 28 at 6pm GMT (21.00 Moscow time) on KGS (English Game Room).
kb: The ranks here are still way off. I'm about Chinese 3 to 3.5 dan strength but that puts me at AGA 6 dan...? I am probably somewhere around AGA 4 dan strength.
Herman Hiddema: A friend of mine went to China as 1 kyu and was, by the end of his three month stay, considered 3-4 dan strength by his pro teacher. Returning to Europe, he had to struggle to make 1 dan, and is currently (1 year later) a strong 2 dan, verging on 3 dan. So perhaps Chinese grades are off? Also, when he returned, his increase in knowledge and in feel for the game was obvious, but it took some time for that gained strength to adjust to the western style of play, so it may be that when you return you perform like an AGA 4d and it takes some time to adjust and reach AGA 6d.
tderz: I quote kb who asked (rethorically) ' I'm about Chinese 3 to 3.5 dan strength but that puts me at AGA 6 dan..?'. It is ambiguous to me what this that here refers to. Any multiple choice test even gives a monkey a certain rank. Common access (Asia, Europe, America) to a unified server might give an answer, however the time differences might not allow the people to play each other. In addition one could differ between online strength and over-the-board strength as the conditions might be totally different (length of games, biological time of play, seriousness etc.).
As EGF 3 dan I had my own experience in a Chinese tournament where I achieved a 3d result e.g. lost vs. 5d, won vs. 2d, won vs 4d, won vs 1k, lost vs. 2d, lost vs. a 4d, etc. My main problem there was the 45min sudden-death time setting.
I met David Wu, a Chinese/French 5d on the tournament, and his opinion (more qualified than mine) was, that the Chinese rankings are 0.5 dan grade weaker than Europeans. This is his opinion and certainly would be verified only around his level.
I only can say that the dans there are shape-strong IMO and some were really playing fast, strong and confident. In the 4 dan's player game; whom I lost against, I thought for some moves that I must do something wrong, so fast and determined came his moves in a certain direction. In the other game with a 4d player, which I won, I almost entered a taisha, pondered and then stopped 'no way' as he might know so many more variations about it, that I created a very simple variation. Later, he entered my moyo skillfully and again with fast and determined play, yet answered (submissively?, goodwillingly, greedy?) my 2 leaning attacks and then I could kill his dragon. However, it was his gentlemanship, that gave me the win, as he stopped the clock and resigned when I only had 1.5 min. left. Thus, time settings might play a very important role. An n+1 dan under 1.5 hours + byo conditions might be only an n-dan under 45m sudden death. It is only human to err and commit so many mistakes.
I had the same experience as another German player, when playing in a park or in a club:
either your opponent or a kibitzer might ask you (for curiosity) for your 证 'zhèng', resp. Dan grade 'certificate'. Well, first, those playing in the parks (perhaps not having, or willing to pay the money for the club fee) don't have a certificate either (they had to laugh then), AND
once you have won your first game nobody asks anymore (this happened only twice to me).
Reflame I think there is an easy way to fix the differences among ratins systems - why do associations not do it this way?
Snoopdogg: Will Sensei's Library change the ranks on the page due to the IGS's rank inflation in March 2007? I'm 5k on IGS currently and I don't know what that converts to AGA.
TapIr: EGF-KGS -> I strongly doubt that EGF ranks are one stone stronger around 1 dan as shown in the list. But don't dare to change it as well.
Tapir: I wrote on the main page that i dislike the way a "5-14kyu band" on Oro is stated. Besides, 5-8d is omitted. (gosensations about Oro: "Even pros have trouble to stay 8d"). Anyone with a proposal for Oro 5-8dan? (Somewhere the "Oro ranks are 2 higher than EGF" seems to change.)
tapir: I've seen several people significantly below their EGF or KGS ratings on OGS. However, those making an entry in the data collection are quite similar to their egf ranks... so, is this systemic bias?
tapir: The internal conversion of KGS ranks is: R(OGS)=2dan + 1,5 * (R(KGS)-2dan). Gnugo being rated around 14k on OGS against around 8k on KGS is pretty much in line with this. Also other data supports this conversion.