Life In 19x19

EdLee wrote:

speedchase wrote:What?

I agree with yoyoma.

You think it's possible that I am actually 18kyu?

speedchase wrote:You think it's possible that I am actually 18kyu?

I think he meant for around 30k levels. (My interpretation of what he said.)
Also, it seems obvious (to me anyway) whenever we try to approximate 100 billion neurons (give or take a few millions)
with ONE SINGLE-DIGIT number, there's bound to be some margin of error.

And no, it's highly unlikely you're 18k, although I would not bet my life if you NEVER make a single 18k move in ANY of your games.
(I know I still do, from time to time. I heard even 9-dan pros do too, albeit very rarely.)

As a relative newcomer to the game, the ranking system seems a little unwieldy (decreasing then increasing numbers around an arbitrary point? 2 uses for ‘dan’?), and only persists due to tradition and inertia. It also seems illogical to be top-down rather than bottom-up. I could imagine a system where the rank of 0 kyu can be set at playing random but legal moves, and then every incremental kyu represents the number of handicap stones needed to match that play, which can be objectively calibrated against such a random player. In practice, this would mean really new human players might start at around 80, while the top pros might reach 150.

Stargoat wrote:hyperpape: Be careful about assuming you know what a person’s real question is.

Huh?

speedchase wrote:

yoyoma wrote:It's accurate to within +/- 15 stones.
Dan ranks are accurate to +/- 3 stones.

Seriously I think it's amazing that this rank system works as well as it does. I think people expect way way too much accuracy from it.

What?

A person who (on average) plays at 1 dan strength might, on his or her best day (well rested, has a good plan in mind for the game going into it, remembers to apply solid principles he/she has learned, playing under comfortable time controls, not worried about errands that he/she must complete later that day, etc, etc, etc) may play at 3d in strength. Likewise, that player on an off day (didn't sleep well, plays without fully considering their move, worried about his/her dog's illness more than the game, etc, etc) might play closer to 3k in strength.

Likewise an average 25k player on his or her best day (doesn't ignore groups in atari, notices an important cutting point his/her opponent has left, plays tenuki when his/her group is already alive, doesn't play little endgame moves too early, etc) might play at 15k level, while on his/her worst day (ignoring a critical 25 stone group in atari) might play much worse than that (let's call it 40k).

speedchase wrote:

EdLee wrote:

speedchase wrote:What?

I agree with yoyoma.

You think it's possible that I am actually 18kyu?

I meant:
30k is actually 30k +/- 15.
1d is actually 1d +/- 3.

3k ranks are more accurate than 15 stones and less accurate than 3 stones.

Besides why not think you could possibly actually be stronger than 3k?

RobertJasiek wrote:The average new player? It must depend on how uninfluenced they remain. Those playing for themselves without meeting experienced players have a much greater chance to stay weak than those receiving at least some external advice and input.

Yes, of course teaching is very important. But I'm not asking how strong people remain after a few hours of good-quality instruction, I'm asking how strong they are as absolute beginners.

I may have confused matters by referring to “one of their very first games” and “3 games out of 3” instead of “their very first game”. This was shorthand way of clarifying that I’m talking about brand new players who nevertheless understand the basic rules.

RobertJasiek wrote:"Please play! I just want to study exactly how weak you remain while I watch and assess it. I do not give you any hints to possibly make you stronger."

Sure, that does sound weird, but scientific experiments are often weird in their efforts to avoid contamination. (In this case, contamination by instruction.) Certainly, anyone expressing interest in the game could be offered real teaching after the experiment was over. And it doesn’t have to be an invasion of foreign homes, it can be willing volunteers.

hyperpape wrote:

Stargoat wrote:hyperpape: Be careful about assuming you know what a person’s real question is.

Huh?

Well, you started your post by saying, “Short answer: yes, there are 30 kyu players.” In other words, you think someone is asking whether anyone is so weak as to be 30 kyu. Who’s asking that? Certainly not me -- I'm convinced that most people start a good deal weaker than 30 kyu. (Look at the last paragraph of post #9.)

The questions I am asking are about the history and accuracy of 30 kyu as an average new player’s rank, as indicated in this thread’s title and first post.

speedchase wrote:

Stargoat wrote:speedchase: I agree with your approach. As you say, testing can be difficult because people’s ranks keep changing. (Here I’m thinking not of the newbies themselves but of the 21k’s we’re comparing them to -- as we go through the process of testing the newbies, the 21k’s get better, which distorts the test.) But if we design a computer program to be 23k (and have it play against a bunch of humans in the 14-16k range to confirm/fine-tune its rank), then we can pit the 23k program against the newbies, and the program won’t get any better. (Unfortunately I don't have the time or resources to carry this out myself...).

in order for this method to be effective, the computer would have to play like a 21kyu, not just play at 21kyu strength. perhaps it would be better to have alot of beginners play, and link it too a known rating system

While I agree with you that the computer’s style of play would have some effect, I tend to think the effect wouldn’t be very large. Can you describe a reason why it would matter a lot? Like, maybe computers in general have greater potential to play effective tenuki than human players of the same rank, which might make it harder for their opponents to keep track of the board?

Stargoat wrote:

hyperpape wrote:

Stargoat wrote:hyperpape: Be careful about assuming you know what a person’s real question is.

Huh?

Well, you started your post by saying, “Short answer: yes, there are 30 kyu players.” In other words, you think someone is asking whether anyone is so weak as to be 30 kyu. Who’s asking that? Certainly not me -- I'm convinced that most people start a good deal weaker than 30 kyu. (Look at the last paragraph of post #9.)

The questions I am asking are about the history and accuracy of 30 kyu as an average new player’s rank, as indicated in this thread’s title and first post.

This is both over-sensitive and a poor reading of my post.

30 kyu assumes a player knows the object and rules of the game and little else. If a player fits that, they're said to be 30k. A player who can give such a player 3 stones comfortably is said to be 27 kyu.

If we change the system and say a player who knows the object and rules and little else is 20 kyu, then we're anchoring the system there. So a player who can give such a player 3 stones would be said to be 17 kyu.

This dictates that a 1 dan under the second system would be weaker than a 1 dan under the system anchored at 30 kyu, since the steps between ranks are theoretically the same under both systems.

That's theoretical. In reality, if you're 25 kyu and on a losing streak against other 25 kyu players, your rank will dip. On a hot streak, the rank will come up a stone or three and the pure confidence boost that gives might keep that rank up while you're learning. And stones in handicaps lower than heavy handicaps mean little until you know how to use them. So, in truth, it's an open system, arbitrarily set with a bottom at 30 kyu which is convenient enough because it's far enough below the level of skill where handicaps begin to make sense. As your level comes up, it's like watching someone new appear out of the fog, bit by bit: hey, it's a new player!

Kyu ranks are a pretty modern thing, right? When did they start? What was their meaning and goal originally?

The dan ranks came about in Japan starting with Go. The strongest player was given the rank of 9 dan, with the handicap system seperating ranks the way amateur dan ranks do today, down to 1 dan, is my understanding. Below that, there really wasn't a ranking system.

This from reading about it on sensei's and wikipedia and such over the years.

I seem to remember that they kyu system did come to go much later ... yep, checking wikipedia it came in the late nineteenth century, here: http://en.wikipedia.org/wiki/kyu.

So, this is a good reason to put the lowest rank to 30 kyu, combined with some ideas in my last post. The strongest players are given a rank ... 9 dan on many online servers, 7 dan in AGA, whatever ... a rank so low that ranks get muddled at that level is a convenient place to put the bottom, but you could put the cut-off anywhere. For instance, on IGS (pandanet) there's a cut-off at 17 kyu or something (it's changed at least once, I think), and anyone below that is ranked BC (beginner's class) or something like that, meaning there's a lot of variability in that class, I would imagine, though I never played there as a beginner.

Yukontodd wrote:The dan ranks came about in Japan starting with Go. The strongest player was given the rank of 9 dan, with the handicap system seperating ranks the way amateur dan ranks do today, down to 1 dan, is my understanding. Below that, there really wasn't a ranking system.

Even though the top rank was 9 dan, the strongest player was not always strong enough. Also, there was at most one 9 dan, so that there were times when some players were nominally 8 dan, although they had 9 dan strength.

The handicap difference began as two ranks difference per handicap stone. Over time, the range tightened, so that now there are about four ranks per stone. (With the end of the oteai system, I don't think that pros take handicaps anymore.)

Let's say we freeze time, and also freeze the brains of every person in the world who knows how to play go. We don't have to worry about any learning, and we have all the time we want. We then have everyone play a 10,000 2-stone handicap games against Lee Sedol. Whoever can win 5000 of those games we declare 7-dan. Then everyone takes 2-stones from that person for 10,000 games. Whoever wins 5000 of those we call 5-dan and so on. Eventually this chain has to halt somewhere: we will find someone who cannot give 2 stones to anyone.

How many steps do we have to go through to get to that person? Will they end up 20 kyu? 30 kyu? 100 kyu?

(There are numerous complications to this scheme: strength isn't fully transitive, you could define ranks with no-komi games or 9-stone games and get a different result, etc. But it's fun to think about anyway)