is KGS 9d a rough ruler for 1p strength?

[Csama]

ok by "1p" i mean the strength needed to become a pro in either japan,korean,china

as i know, KGS is the hardest server to get 9d
all the 9ds are mostly pros (milan,smartrobot) or ex-insei(korondo,shawn)
since korondo beat dinerschtein and catalin in EGC, his strength is also like 1p
even shikshin, who plays dinerschtein and catalin evenly, is still 8d

and i read a chinese report before,
before Joanne Missingham aka Hei Jia Jia became a pro,
she asked her mom to take her to China pro exam if she can achieve 9d on the internet
she didn't say which server but i guess it's KGS
and she indeed became a pro!

so what do you guys think?

[Solomon]

If you're beating players like them, sure. If you're one of those blitz hogs who only play 6h games against 3ds hoping for time wins, then no.

[Joaz Banbeck]

If I were, say, about 7D strength, and really wanted to have that 9 after my name, I would have multiple accounts. I would play each of them, and the laws of probability would dictate a bell shaped curve of results. Most of my accounts would be 7D or 6D or 8D. But one would occasionaly sink to 5D, and one would rise to 9D.
I would abandon the 5D account. And I would save the 9D account for boasting.

So I suspect that many of the seldom-used 9D accounts are owned by players who play regularly under other names, and who could not maintain that rank if they tried to.

[lovely]

I think 9d on KGS could be a low-dan professional in Japan, and ex-insei/6d amateurs in Korea and China (except a handful who are actually professional).

[emeraldemon]

Well, there's only 14 9d on KGS as of this writing, so it's easy enough to see how much they play:

Code: Select all

name        games   july
jove        27  3
eku         53  2
evildoer    18  1
ericdai     96  2
maynerd     28  15
korondo     88  14
mysterious  99  3
chijoong    9   9
milanmilan  657 33
smartrobot  213 11
shawn5475   32  3
stc         6   0
lsai        55  7
pozz        7   7

That's total games, followed by games this July. Not sure what counts as "active", but only milanmilan and smartrobot have played more than 100 games on the server (although mysterious is quite close). Three of the players have fewer than ten total games, so I expect rank to drop for them. Overall it seems a very difficult task to achieve. I wonder if it's possible to see the 9d list for 1 year ago, for example? How many of these names would still be on it?

[Oroth]

Not sure you can make a meaningful comparison:

Some 1p's are stronger than 9p's, so what is 1p strength?
9d is the limit on KGS, so if Lee Sedol played on KGS he would be 9d, so what is 9d strength?.

If you are asking whether a player who made it to 9d would be strong enough to make 1p. Possibly, but of course making 1p would be far far harder. There are many many strong amatuers and inseis, capable of regularly beating pros (say nearly half their games), who would be 9d on KGS, but wouldn't make it to 1p because there are only a couple of places available a year.

[Kirby]

To be a pro, I think that you need to be at least good enough to be KGS 9d. But I think that, even if you get to be KGS 9d, it might not be quite enough.

There are examples of old pros, I think, that have 8d accounts, for example, but I think that if any of today's pros buckled down and focused on winning, they would be 9d (if they ever played on KGS).

So if you are trying to become pro, first become 9d. After that, get even stronger.


---

Aside from that, I agree that the comparison is not totally meaningful, because you are measuring different things. Being a pro is like having earned a certification, whereas KGS 9d is just a result of winning games.

The discussion I have above corresponds to my opinion of the strength of those that currently are able to achieve professional "certification".

[quantumf]

KGS has the strongest or most severe ranks of the well known servers. The genuine 9d players (i.e. who can maintain it over a series of many games) are probably of pro-like strength, which is not the same as being good enough to become pro. Achieving this level on an internet server, where there's nothing at stake, is very different to beating out other similarly strong players to make the grade in the pro exams.

As an illustration of the difference between servers at this level, Roseduke (also a former insei), holds a 60-40 win ratio as a 9d on Tygem, but is more of an 8d than a 9d on KGS.

[hyperpape]

Not sure you can make a meaningful comparison:

Some 1p's are stronger than 9p's, so what is 1p strength?
9d is the limit on KGS, so if Lee Sedol played on KGS he would be 9d, so what is 9d strength?.

This is not right. The server computes a rating that extends up past 9d. The server only displays up to 9d, but the rank graph will show you how much a player exceeds 9d. Some professionals have maintained 10d strength on KGS for a period of time (under the new ranking system).

That last fact makes me hesitant to say 9d KGS = 1p. There's internal variation in the strength of the KGS 9ds. Bigbadwolf wasn't winning half of his games against MilanMilan, afaik.

[jts]

I thought that the re-centering was constantly compressing the ranks, so that the highest rank is 9d.

[hyperpape]

Maybe. The examples I looked at were several months old. When was recentering implemented?

[jts]

I believe recentering happens every time the ranks are reiterated (i.e., continually). I'm not entirely sure how the math works, but the point of running the ranking algorithms with anchors is to keep all ranks between 30k and 9d.

The KGS rank system uses an anchor system. It is just a final translation on all the computed ranks to ensure the ranks displayed are around 30k to 9d.

So (if I understand what this means), the reason that e.g. MilanMilan appears to get much stronger (what would be "10d") and then suddenly gets much weaker over the course of a week (normal 9d) isn't that he's suddenly losing more games. Instead, the algorithms are designed to compress the ranks of all active players between 30k and 9d.

[mw42]

At least jove is currently rated over "10d."

See: http://www.gokgs.com/graphPage.jsp?user=jove

[wms]

KGS doesn't compress ranks when it adjusts the baseline. It does check after each iteration of the rank system, and it will shift all ranks up or down by a small amount, but it doesn't try to keep the strong players in 9d. Instead it has some reference players whose rank it "knows" (called anchors) and it adjusts the whole population until the average anchor rank is equal to the average "known" rank of the anchors.

[daniel_the_smith]

Do the anchors know they are anchors?