Life In 19x19

If your anchor players (e.g. Zen) could play enough games on the go server, then I speculate that eventually the ranking algorithm wouldn't matter. It would be an interesting mathematical challenge to figure out how much "enough" is and how "long" eventually is, for any given server and it's algorithm.

quantumf wrote: If your anchor players (e.g. Zen) could play enough games on the go server, then I speculate that eventually the ranking algorithm wouldn't matter. It would be an interesting mathematical challenge to figure out how much "enough" is and how "long" eventually is, for any given server and it's algorithm.

For this to work I think the players on both servers would need to be the same (and require that they interact with the anchor players at a high enough frequency and in the same way).

In a sense, the anchor players are part of the algorithm. We can think of them acting a little like interpolation points, while the rules for ranking up act like the interpolation. But this is an imprecise analogy, as their influence on the rank-curve would depend on their interactions with the players on the server: both the frequency of interaction (how often does a typical player play against an anchor), how do they play against the anchor (do they always play even or do they more often play with a handicap) and how well do players on the server play against their style (do they do well at fighting, but less well with large moyos).

S2W wrote:[good points...] In a sense, the anchor players are part of the algorithm. [more good points...]

Yep, all true. An additional point which you indirectly raised is that there's a reasonable likelihood the anchor players will have different ranks on the different systems. Even if you had the intention to fix their ranks, algorithms like the one KGS uses allow for re-anchoring and subsequent auto-adjustment of everyone else, but I'm pretty sure the Tygem algorithm does not.

If I remember correctly, your rank on Tygem is based on your 20 last games: if you win or lose 14 games out of them, then you rank up or down. Then, if you had a winning/losing strike that goes on for a few more games (how many I'm not sure), you'll continue ranking down/up pretty damn soon. But if the strike stops quickly after changing ranks, then you'll have to complete your new 20 games set on this new rank (with 14 wins/losses out of 20 games) before going up or down again...
I like that you always know how many more wins/losses you need to change your rank. I don't see having different ranks on different servers as a problem: I know I'm 1dan on Tygem and WBaduk but only 3kyu on KGS, so I feel 1dan players on Tygem are nothing special, while they frighten me on KGS...
The only issue I can think of with all that would be at the utmost rank: there seems to have quite a gap in strength between some 9dan players on Tygem (just look at Haylee's winning ratio there). Which means that they could either promote the stronger players to 10dan, or keep them to 9dan and demote all others to a lower rank (except for beginners?).
But then I wouldn't be a 1dan on Tygem anymore

I don't play that often, but I've been a 1D on KGS for a bit and just recently ranked up to 3D tygem, and the last few games I played feel even. Might matter what time we're playing, too, given Tygem's quirky rating "system." Based on my results from people who had several hundred+ games, versus those with far fewer games, I could be convinced that if I don't get stronger I'll stabilize around 4D on tygem.

erislover wrote:I don't play that often, but I've been a 1D on KGS for a bit and just recently ranked up to 3D tygem, and the last few games I played feel even. Might matter what time we're playing, too, given Tygem's quirky rating "system." Based on my results from people who had several hundred+ games, versus those with far fewer games, I could be convinced that if I don't get stronger I'll stabilize around 4D on tygem.

I can second that.