So I am currently ranked at 2 kyu on KGS. However, I hopped on my old Tygem account (where I have held shodan for quite a while and ranked up to 2D easily, and now at this level my record is five wins with only one loss. So I'm fairly sure I will get to 3D. I've always heard that Tygem is a 2 stone difference from KGS but FOUR stones? This blows my mind. I've realized that Tygem players, while loving to fight, don't pay attention to basic tesuji and single digit kyu life and death. Has anyone else seen or felt such a difference in level of reading on Tygem? Have you guys experienced or heard of such huge rank discrepancies? (sp?)

Yes. 4 stones is consistent with what I see. In some cases, it might be more.

So why the huge difference in strength?

the strength is the same, the numbers are different.

I am a KGS 3-4Kyu. I rarely play on Tygem, but my account is at 9K, and when I do play there I end up losing about half of my games. When I lose it is usually because of a misread after a huge fight that resulted from continuous unreasonable aggressive moves. My own experience is that the players at 9k on Tygem have virtually zero strategic knowledge, but are pretty decent readers and fighters. I tend to avoid playing there because I feel like less of a player losing at 9k. It is probably a really good way to improve my reading though.

However I've met some players on KGS the same rank as myself who do much worse on tygem? so why do some people do much better, other far worse?

I am 1D on KGS and 5D on Tygem. There are servers even worse, where I know people of my strength to be 8D/7D's. Basically your rank means nothing, it's only value is within the referred group. One day I would like a standardized computer programm like Zen, Aya & Co. to be a kind of benchmark, playing on all servers and making sure all ranks are about the same. I have no idea about the math behind it, but the idea is to have a fixed strength called S (defining it as "1d" for example) and then evaluate all players based on their performances directly or indirectly based on S. As it is now, one must be careful, because if some dude tells you he's 8D it could be a 1k in fact^^.

So if you create some sort of benchmark, that benchmark will have to select some model on which to base its rank calculation. It could be similar to KGS, taking into account recent games, for example. Or it might be more like Tygem, taking into account only the results of the last 20 games you played. Or it could be a new system altogether.

But selecting such a baseline would simply give you a rank for people according to the model that you select - if your model for the baseline ranking system is closer to KGS's algorithm, the ranks will have a distribution similar to KGS ranks; if your model for the baseline ranking system is closer to Tygem's, the ranks will have a distribution similar to Tygem's.

I would recommend you start a new account and set your rank to 3k and see how you do from there. At the lower ranks there are a lot of sandbaggers.

I started a new account just now and won my first game as 1 dan. Ain't that a kick.

I do not believe that, though I have nothing to back it up . KGS and Tygem - even with their different models - would get a lot closer if they have a common benchmark. That'd be my guess.

Try to make a new one at 3d and you may win that as well.

How's your game record on that 5d account you made?

Unfortunately just the one game so far still.

Well, I guess once you play more games, you can let us know if the strengths are as random as you suggest.

Also since handicap games aren't rated games there, I don't think ranks are at all stone differences, so that also makes it hard to do any direct comparison where KGS has quite a number of ranked handicap games.