Here is my stats:

I'm 5-6k on kgs and about the same on tygem.

I'll probably get harder games around 10-18k though but I guess we all know the reason for that

Cheers,
Otenki

There's a 'social' reason behind the Tygem sandbaggers.
In China, NO serious player should have a starting rank below 5k-1d.

Serious player = You're a kid and your parents have a teacher coming once a week or you're a little older, you played at school and took some lessons.

So if you're a 2k KGS and 2d Tygem it means you played against kids whose teacher told them to register as dan level.
(that's why you hear a lot of people saying Tygem dans are weak)

In the ddk range you will only find people messing around and trying stuff. They are not sanbagging to get easy wins like you find in the west. They just try stuff or play randomly because they are bored.
You can find anything: the real ddk who never played or the high dan messing around.

So for a KGS player who wants to play on Tygem:

-If you like to play at your own level, register at a higher level than your kgs rank, even start at 1d if you're in the sdk range.

-If you like to play strong &| tricky players, register at 20k-10k (or even lower). You will get some garbage but eventually you will play against a strong player.

No amount of statistics will help create a rank comparison for Asian servers.

Actually it's the same for new go servers like Kaya or Nova. I don't know why sdk like to register as 30k

PS: If you like to know my stats, I'm 7k KGS and 15K Tygem (started at 10K and I'm going down steadily)

I'm Tygem 6d, but that's not really solid as I'm on 20 wins out of 22 games since 5d, 5 wins of 6 since 6d. I was KGS 4d or so, but maybe I've got stronger? My mindless KGS blitz account is 4d not 3d now.

No amount of statistics will help create a rank comparison for servers that allow players to choose their own rank.

Since handicap games are not used in the rating systems for cyberoro/tygem, I would expect there to be quite a lot of difference from kgs/igs.

What I battle to understand is how the discrepancy between Tygem and KGS ranks seems arbitrary for some players, and not for others. I'm 10k on Tygem, and 5 or 6k on KGS. Several other people have noticed this difference. However, an almost equal number of people report their ranks as being similar across both servers.

How can that happen?

The ranking systems are fundamentally different. If a similar type of ranking methodology was used between two servers with different groups of people, maybe you could compare them. But the systems are very different.

Like people have mentioned, you can specify your starting rank (up to a certain level) on Tygem, which leads to variation particularly for people not having played many games.

In addition, Tygem's system is not as sensitive to history.

_________________
be immersed

That still doesn't account for the people who play a reasonable amount of games on both servers, such as the posters here. How do some have the same rank on both servers, and others are vastly different?

In theory, as the number of games increases, the ranking systems should become more similar. However, there exists the phenomenon in real life, aside from go servers, where player A consistently beats player B, player B consistently beats player C, and player C consistently beats player A. My hypothesis is that this could be due to style. Perhaps player A plays in a style that player B is weak to, player B plays in a style C is weak to, and so on.

If style can contribute to such a situation, then I don't find it unreasonable that Tygem and KGS would as a whole have different styles, and therefore you can see cases where we have players of "type A" who can consistently beat the dominant population of "type B" players playing on Tygem, but also have players of "type C" who can consistently lose to the same group of people.

Theoretically, the distribution of player types, A, B, and C, could vary between tygem and KGS. KGS could, for example, have 50% type A players and 50% type C players, with maybe Tygem having, theoretically, 95% type B players. In such a case, when the KGS players play on tygem (obviously contributing to the 5% that aren't B players), then we would have half of the KGS population consistently beating Tygem players, and half consistently losing to them.

These player types, A, B, and C, are of just made up, and we have no idea of the distributions of "player types" on the servers. But it is a hypothetical explanation for the behavior that we see here.

_________________
be immersed

Tygem's rating system isn't designed to match equally strong players, just to create some amusing noise.



Hypothesis aside, do you have an evidence of actual triplets of players with large numbers of plays who exhibit a dominance cycle like this?

I agree. I've found that tygem players, at least the ones I've played, seem to not put as much on whole board thinking and strategy as KGS players do and that they end up just fighting and invading. With KGS players, I've found the opposite to be true. Of course, I've seen some exceptions, but by and large I've found the aforementioned description of each server's players to be true. Therefore, I think that if you're decent at direction of play and whole board thinking, but really good at reading and fighting, then your tygem rank might be higher than your KGS rank, with the opposite being true if your specialties are reversed. Hope what I said made sense .

I am a pretty stable 1d at KGS. Yestersay I played my first game against a 1d-Tygem...and felt like playing a 2-3k at KGS. Even worse: I remember dwyrin/battousai/renshai losing to a 2k and (almost to) a 1d at KGS (he didn't because the 1d played a blunder at the end) 1 year ago and he was/is 5D there at Tygem. IMO there is a definite rank shift between KGS and Tygem, even so gross that I'd find it annoying. Why don't those servers just use bots with a specific hardware config. as anchors?

That sounds fairly accurate.

I'm KGS 1k/1d and 3d on both tygem and wbaduk, though I don't play much lately.



There aren't yet strong bots available to have anchors across all ranks. Also people may play less serious against bots, or may refuse to play them at all.
(For example, many claim KGS bots and people who play them exclusively have inflated ratings.)

How does that solve it when every server will think they have the best starting point?

..

My - naive - idea would be like: Take Crazystone 2013 on a computer with hardware x. Then let it play regularly on IGS, KGS and Tygem and use ONLY this program as an anchor. Simplified: Who wins against CS would be rated higher as CS and so would who does win against someone who won against CS; people that lose against CS or lose against people that lost to CS would ranked lower and so on (that's just a simplification to show my point). With a system like that I could imagine a system with consistent ranks between servers. Again: I'm a complete greenhorn here, because I'm not a programmer or mathematician, so maybe there are big flaws in my idea. I just would like the idea of a unified rank system and computer programs seem a good way to do that - at least for me.

I don't see it happening for a variety of reasons. First, the major go servers (Tygem/cyberoro) don't have an interface I know of for adding bots. The second is that the systems and how they account for handicap or rank differences are very varied.

It's easier to just play and let the rank settle it out to whatever. The only point of the rank is to get relatively even games on each server. They can be wildly different but there's no reason to care.

The big flaw in your idea: What is the correct rank for Crazystone? KGS will say X, Tygem will say Y and guess what, they are both right.

My thinking is: Because CS will be played on a fixed hardware and time setting on all go servers, it will have a certain equal strength - everywhere. If one calls this strength "1d" and uses it as an anchor then all servers should get the same ranking.

Hm, I don't really care as long as the ranking system does its job: match players of approx. equal strength against each other.
Go strength isn't a 1-dimensional quantity anyway, so I'd try not to attach too much meaning to a single number.

Your idea seems to bring with it a lot of problems while fixing none. (Ranking systems don't have to be comparable, just consistent within themselves.)