Life In 19x19

I'm not really fixated on rank... I was rated in 1976 at 14 kyu and, to a first approximation that seems OK..

I've played a fair few games now.. on OGS, Tygem, IGS (not KGS that much). My friendly 3rd Dan has gone from "have you ever thought about playing chess instead of go?" to "learn this, learn this, do this"... so I guess I've picked up a little. On OGS I believe I have the highest loss to win ratio there..


OGS is great - a tight knit community who discuss stuff. Nice software.. I get through about 1.3 games every day but.. quite clearly because there are few regulars we are all bouncing off each other. Provisional players put in their KGS ranks.. get roasted and herein lies a little of an issue. On a number of occasions I've beaten EGF 4Kyus, AGA 5Kyus the odd KGS 2kyu with little/no handicap. at the other extreme on Tygem I get wiped out by 18 Kyus (OK I know its different time scales). At lower levels the ranks are out of sync..

The question.. has there been any move by, for example the IGF to define global ranks. In the past this would have been a no no.. but now with the internet it must be very close to possible.

I am also aware that in Go, ranking is largely determined by playing. In say, martial arts, your rank is usually determined by a combination of tournament performance and underlying knowledge e.g Kata. This latter feature seems to be missing in Go and is sad because you might get a situation where high Kyus know how to win but don't necessarily know core knowledge/history which might enrich their experience and those who learn off them.

As in martial arts, or driving tests.. there might be some merit in having exams in Go as well as a ranking event/tournament.

Whats the point?.. Well rank obsession seems to be a feature of DDK and SDK's - feeding that obsession with a good benchmark might promote the game. I have no experience of Dan grades but it seems that at that level each grade is awarded in a more considered way over a greater amount of time.

It is possible (what isn't?), but it is far from being realistic.
KGS, WBaduk, Tygem, IGS, OGS, to name the most frequent online Go servers, all have different ranking systems/algorithms. Then there are the national (AGA [USA], Japan, China, Korea) or EGF ratings, which again are all very different.

I know players, who play a five Dan on Tygem, a two Dan on KGS and have a one Dan EGF rating.
I am three Kyu on KGS but I have a six Kyu EGF rating.

So the one thing is to overcome all these differences and adjust every Go server's and national (+EGF) algorithm. The second thing is, which system is the most favourable/reliable/sensible and should therefor be promoted?
The third question would be: What do we gain with worldwide comparable ranks?

A universal rank is hardly likely o be acheived in the near future.. if the evidence of martial arts is anything to go by.

A universal benchmark is different.

It is straightforward to create online tests and normalise these tests to appointed players.
Appointed players might also play a range of servers. After time, the tests would hopefully converge. To get a "formal benchmarked rank" a player might play a server of choice for a few days and do a few days of tests. At that point, you may have a reliable rank..

But.. whats the point? As with all benchmarks and standards they are there to improve a system. Improvement could come about by making sure new players have minimum knowledge before they progress (e.g know basic joseki) and at higher grades ensure newer ideas are injected into the game through the players.

Go has a large teaching component.. a high grade might mean that certain competances have been met and new players would be assured that a low SDK actually has a breadth of knowledge about the game as well as being able to play well..

Possibly this is something the IGF should work towards - it does seem to fit into their remit.

I would point out that at least one Korean University offers degrees in Baduk. Those degrees will be benchmarked in some ways.

It depends I suppose if you consider ranking to be a tool for simply handicapping games or an instrument of furthering the game.

Some people will play in a game a lot better than they will perform on a test, and vice versa. Since I believe rank is meant to be an approximation of your playing strength, I think basing it on anything else would make it less valuable. Could you explain what you mean when you say rank is "an instrument of furthering the game"?

instrument of furthering the game:

Well to an extent it already is. There are examples of ranks being awarded for reasons other than strong play.


Strong play in itself doesn't make someone good at teaching ( those who can do.. those who can't teach). There might be some merit in promoting an aspect of the game that recognising the "academic" work that is put in and the value that people can add to the game other than beating their immediate peers. It would also help new players as they would have some guarantee that if they are being taught by a 5kyu.. that will get a base level of insight into the game (other than - well this moves works pretty well).

You do see examples of people putting in a lot of time and effort into reaching certain grades. The extent of that study probably doesn't reflect the final reward and of course if you can claim you are "shodan" AGA there's always the "well thats not a real Shodan is it?" kind of effect.

I suppose that at my lowly levels I see a lot of fluctuation in games - to the extent that I'm not sure I even have a meaningful rank. I'm thinking of going off to an open tournament.. just for fun... but I have no clue what rank to declare.. If there were online exams run every three months or so combined with a portfolio of games played over a few servers (which had been reviewed) - at least I'd have a fair stab.. Probably I'd be prepared to pay for it.. if I had a pretty certificate at the end...

All this would cost time and effort but surely the IGF must have made some sort of move in this direction?

While I agree that there is some benefit in ranking systems having the same scale everywhere, that's hard to do in practice.
And why do you think a rank based on some arbitrary trivia knowledge is any better than a rank based on playing strength? Why do you expect a kyu to know any joseki?

Unifying the AGA and EGF system should be possible. The main problem here is that there are not many games between these two groups of people.
Tygem has no working ranking in the kyu ranks at all. So no amount unification will help there.

Online vs. offline, turn based vs. realtime, blitz vs. slow can shift the relative strengths of players by several stones. I'm sure I can beat some players on kgs with several stones of handicap who need much less handicap offline.

Li Kao wrote:While I agree that there is some benefit in ranking systems having the same scale everywhere, that's hard to do in practice.
And why do you think a rank based on some arbitrary trivia knowledge is any better than a rank based on playing strength? Why do you expect a kyu to know any joseki?

Unifying the AGA and EGF system should be possible. The main problem here is that there are not many games between these two groups of people.
Tygem has no working ranking in the kyu ranks at all. So no amount unification will help there.

Online vs. offline, turn based vs. realtime, blitz vs. slow can shift the relative strengths of players by several stones. I'm sure I can beat some players on kgs with several stones of handicap who need much less handicap offline.

Well.. it certainly wouldn't be up to me to decide minimum standards for any rank... but even now there are numerous tasks that are useful for developing go technique (problems, game guessing, joseki) that lend themselves to assessment. I find it hard to believe that anyone at around 10 kyu doesn't know 3,3 invasion, approach moves and basic fuseki. Certainly at my lowly levels having a formal structure of what is important to know (and knowing you might be tested on it) may serve as a catalyst for more structured (and faster) learning. Currently the advice at lower levels seems to be:

1] Go lose 100 games
2] do some basic problems.

In order to get to shodan the advice seems to be

1] get your games reviewed
2] play >= 1000 games
3] do more problems faster
4] learn some joseki..but not too many or else you'll drop two stones.

The bit to shodan I gleaned from here.

The guy who teaches me propounds the theory that I should only give Dan players two stones (irrepective of the Dan rank) and that if I happen to play a SDK I should setup a small group for him to attack while I run off and make territory... In that way I'll make shodan in 2 years...

For a game of such high brow roots.. it does seem a little derelict in a structured approach for newbies..

Other thoughts..

In Judo/Karate for example.. each rank requires set moves to be demonstrated. These are basic moves that must be got right before a higher rank is won.


Actually.. I quite like the tygem ( rank judged on last 15 games) and Wbaduk (play alot) ranking systems.. they seem more mature.. but that's just me...

I'm not really interested only in the extent of handicap. The purpose of rank could also feed teaching.. and at the very high levles make sure new ideas are fed into the game.. e.g some Go "body" might reflect on the current state of the game and decide that in order to reach shodan certain NEW ideas / constructs should be understood. In this way the game may evolve..

Isn't the sole purpose of rank to set the appropriate handicap (ie the difference in rank is the handicap for an evenly-matched game)?

Recognising other accomplishments and contributions is very nice and all. But you could do it with badges, certificates or whatnot. We would still need a means of ensuring players of different strength could get a good game against each other.

i don't think that this idea, while sounding interesting, has a chance of being realized. certainly you can't do such exams for every rank, but let say every five ranks (15k, 10k, 5k, 1d).

but
1. who would decide which knowledge makes n-th kyu and who would force national organizations to follow this new standard? i don't believe IGF would be capable of that, not even EGF in Europe. we have one European rating, but rank awarding still differs between some countries

2. who would run these tests? if you propose to run them online, you have virtually no way to avoid cheating. if you don't care about it, this rank would be only for orientation, being less reliable than current rankings based on game performance. if you were to run the tests in real life, it creates much more money-, time- and space- expenses

3. who would be interested in taking these tests? i think most players would be only annoyed if they had to take tests to achieve a rank, they have enough exams at school etc., they want to just have fun and play some go, not speaking about having to pay for the tests. the comparison with martial arts is tempting, but in fact the situation is different. few go players are in a dojo, where they would have a sensei, who teaches them practically all they know and who would be capable of judging if they picked it up right. go players play and study at home, play friendly games at club or more competitive ones at tournaments and if they are good enough, they win. and i wouldn't like a situation like "too bad i beat the local 8k three times from five, i again messed up the 9k exams, so i still have to play the 10k noobs at tournaments. stupid rank system"

one thing i like about your proposal is that it would make players study more and get better, because they would need the knowledge to pass the exams and achieve the rank. but they would have to care enough to bother. maybe they wouldn't and the system would end up less reliable than current ones

Laman wrote:i don't think that this idea, while sounding interesting, has a chance of being realized. certainly you can't do such exams for every rank, but let say every five ranks (15k, 10k, 5k, 1d).

but
1. who would decide which knowledge makes n-th kyu and who would force national organizations to follow this new standard? i don't believe IGF would be capable of that, not even EGF in Europe. we have one European rating, but rank awarding still differs between some countries

2. who would run these tests? if you propose to run them online, you have virtually no way to avoid cheating. if you don't care about it, this rank would be only for orientation, being less reliable than current rankings based on game performance. if you were to run the tests in real life, it creates much more money-, time- and space- expenses

3. who would be interested in taking these tests? i think most players would be only annoyed if they had to take tests to achieve a rank, they have enough exams at school etc., they want to just have fun and play some go, not speaking about having to pay for the tests. the comparison with martial arts is tempting, but in fact the situation is different. few go players are in a dojo, where they would have a sensei, who teaches them practically all they know and who would be capable of judging if they picked it up right. go players play and study at home, play friendly games at club or more competitive ones at tournaments and if they are good enough, they win. and i wouldn't like a situation like "too bad i beat the local 8k three times from five, i again messed up the 9k exams, so i still have to play the 10k noobs at tournaments. stupid rank system"

one thing i like about your proposal is that it would make players study more and get better, because they would need the knowledge to pass the exams and achieve the rank. but they would have to care enough to bother. maybe they wouldn't and the system would end up less reliable than current ones

1. Testing is a lot more sophisticated nowadays... Bands of ranks is a good idea.. i.e you score 90% you go to 6kyu..you score 80% go to 7 ku --etc.. tests would only seem any good for bands as you say (e.g 10 kyu to 5 kyu).

Who validates?.. This is why I suggested that you appoint players of different grades to play and do the tests. The benchmarking players would all validate the tests and questions by doing them. e.g 5 players of around 7 kyu all do the tests..they identify the questions that differentiate (by getting them right or wrong) Those good questions carry through and are used later.

2. Who runs it? This is price/value. If people who run tournaments begin to recognize this approach as useful.. and use it for ranking -then the tests might be paid for - As per the cartoon above..

The tests are not proposed the be all and end all of ranking. A portfolio of games played on different servers might be submitted for review as well. My guess is that many dan ranked players could grade a player by looking at their game. Test might include a candidate actually reviewing a game... (a task that often occurs out of fun - minumum standards might be helped here).

Would it pay?.. I don't know.. Universities in Korea fund entire degrees on Baduk.. there seems to be money there...

You can never stop a concerted cheat- but you can render cheating worthless. I could almost certainly reach Shodan on KGS in a week by cheating... but there's not a great deal of point.

Who would pay?.. well again the martial arts analogy. People pay for gradings. Go rankings as they stand seem to have little value. Perhaps if people paid for them - got a certificate then the system would improve and have more value.

Last point - you can't force people to care..or even take part. What you might say is that this is the most through assessment regime available that promotes understanding and teaching of Go.. it may not be perfect but it can be improved and it is better than that which went before....

I believe it's possible to have an overall ranking system, but it would certainly require a dozen organizations to desire that.
If the major online servers got together to have a standised ranking (a single rank), offline tournaments would quickly follow. Yes, some people play better offline than online, etc. but no system is perfect and this would be a good compromise.

Well, if the major online servers decided to do that, we would be closer to a centralized gaming platform with different clients (guis). It would be perfect.

This idea does not make sense to me. In principle, someone of rank n should be able to beat someone of rank n-1 about 2/3 of the time. This is expected to (and of course does) have massive error margins, such as pairs where one always beats the other but loses more to everyone else, but it's the only thing that rank is really for.

I don't see why any kind of test system would be worthwhile. I'm not 2 dan because I can solve certain tsumego or perform certain arbitrary tasks that other 2 dans happened to be able to do. I'm 2 dan because, in equivalent ranking systems, it gives a meaningful value for the handicap I should give others.

It isn't clear to me what a test system would gain, or even expect to gain.

amnal wrote:This idea does not make sense to me. In principle, someone of rank n should be able to beat someone of rank n-1 about 2/3 of the time. This is expected to (and of course does) have massive error margins, such as pairs where one always beats the other but loses more to everyone else, but it's the only thing that rank is really for.

I don't see why any kind of test system would be worthwhile. I'm not 2 dan because I can solve certain tsumego or perform certain arbitrary tasks that other 2 dans happened to be able to do. I'm 2 dan because, in equivalent ranking systems, it gives a meaningful value for the handicap I should give others.

It isn't clear to me what a test system would gain, or even expect to gain.

Beating other players should be one measure of rank. I don't disagree.

At my point there is a wide variation of advice on how to improve. In a subject as wide as go there must be priority areas to grasp. This makes learning more efficient. eg There seems little point is a 20 kyu learning 3P games..

If you were teaching someone to be a medical doctor you wouldn't rank a doctors progress by his success/failiure to cure people. It is usual to give advice and guidance for minimum standards of knowledge. It would be a very inefficient way of teaching a doctor to rely on trial and error. Imagine medical students looking at the end of year results..."oh geeze I killed 51% of my patients this year..looks like I'll have to redo the year"


From your point of view you may have insight.. You may be very clear what your priorities areas are for improvement to get to 3rd Dan and lay foundations for 4th Dan.. I would suggest that simply playing stronger players might be an inefficient way of progressing on its own.

Tell me, what additional insight do you need to reach 3rd Dan?

It would be a very inefficient way of teaching a doctor to rely on trial and error.

Maybe, but it's the most efficient way for a go-beginner to learn.

You have to keep in mind, that go ist practicly "unsolved". The best computers can't beat pros or top amateuers on even level, and there is no definite set of knowledge that one absolutly has to know to become x dan.
All we have are various areas of theory how stones on the board interact and how the fact that each player only plays one move at a time, impact the course of the game.

Because every game is different, and all theory has to be applied to the specific game you are currently playing, experience is utterly important. Experience tells you where the key areas are, where you have to read deeper. Experience tells you where the pitfalls may lie and where everything is safe.

You can become a doctor by learning certain facts about the human body and how certain illnesses influence its normal behaivor. But you can't become a good go player by learning certain facts about the game, you can only become a good go player by playing and gaining experience.

Another thing, why a test for a specific rank may not work is, that each player has his own strength and weaknesses. Because there are a lot of areas where one can study, players of the same level (meaning they each have 50% chance of winning against each other), can have varying abilites in reading, whole board judgment, tesuji, life&death, shape, joseki-knowledge, sabaki, efficient use of thickness etc..
Because of this, the results of a test may greatly vary and it would be unwise to use such a unreliable source of information for ranking purposes.

Of course there are certain areas which are more important than other, but it is possible to become a 1-dan with a very high skill in reading and a very faint grasp of certain strategic concepts. On the other another player may not be able to read as deep as the former, but has a good grasp of strategy and therefore find himself more often in favourable posistions, where he don't has to read as deep as the former one, to get a good result.
In the end, both players may be of equal strength.

Tell me, what additional insight do you need to reach 3rd Dan?

From a strong kyu point of view (i'm 2k), there isn't that much additional insight. Of course there are certain things, which strong players use in their reasoning, that weaker players (like DDK) don't even think about. Whole board judgment, direction of play, gaining/keeping sente come to mind.
But if you ask what is the difference between a 3dan and a 1dan for example?
They both know the same things (well maybe, the 3dan knows some more josekis than the 1dan) but the 3dan can adapt his knowledge better to the current game, and he is faster, which means he has more time to read, and can probably read a bit deeper than the 1dan.

In the end, it's not about the amount of knowledge, it's about how well can you use the knowledge.

Also, what areas one has to study to improve depends on the person. Mostly you have to study the areas of the game, in which you are currently lacking. Probably the only thing, that is true for all ranks is reading.
You can always improve by improving your reading.