v00d00 wrote:I notice one consistent pattern on KGS: from summer (around may) ranks of many kyu-players (6-1k) slowly going down. For example: I check 15-20 accounts. ≈80% have this regression.
It's my incorrect statistics or something happens with ranking system (or players? O_o)
I'd noticed something similar, I don't think it's just the 1k-6k range (I looked into it a bit when twoeye dropped to 6d, also TheCaptain has a downward trend which is a bit abnormal). There was a big shift in March upward, and another shift back down in May, likely anchor changes. Another possibility (that I'll admit I haven't looked into the numbers on) is that recently I feel like more strong players have been making accounts and playing on KGS, that could be pushing the rest of the server down.
I went in the opposite direction, but recently there was a sharp jump up followed by a sharp jump down of almost a stone. So maybe there was an anchor change very recently? I haven't been playing on KGS so my rank is very volatile though.
I go through long periods of not playing though, so if there was a downward trend around 10-8k recently it should be visible no?
Boidhre, your rank is really light, so I don't see any unusual things: I think sharp jump down happens because information about your previous win-streak is gone (45 days half-life of the games)
v00d00 wrote:Boidhre, your rank is really light, so I don't see any unusual things: I think sharp jump down happens because information about your previous win-streak is gone (45 days half-life of the games)
v00d00 wrote:Boidhre, your rank is really light, so I don't see any unusual things: I think sharp jump down happens because information about your previous win-streak is gone (45 days half-life of the games)
It doesn't work like that. 45 days is the half life of the games, so it would slowly lose it's value over time. it wouldn't cause a dramatic downward spike.
speedchase wrote:It doesn't work like that. 45 days is the half life of the games, so it would slowly lose it's value over time. it wouldn't cause a dramatic downward spike.
v00d00 wrote:Boidhre, your rank is really light, so I don't see any unusual things: I think sharp jump down happens because information about your previous win-streak is gone (45 days half-life of the games)
It doesn't work like that. 45 days is the half life of the games, so it would slowly lose it's value over time. it wouldn't cause a dramatic downward spike.
Indeed in this Boidhre's case it's much more simple -- The sharp jump up was because after not playing many ranked games he played 5 in a row in mid-august and did well. The sharp jump down is because of those 5 games he played in mid-august, 2 were against a player who hasn't played anymore, one was against a person whose rank is fairly steady, and two were against bentw, a player who didn't play rated for the month of September, but then suddenly played a lot of rated games mid-October and saw his/her rank jump around --
Bentw's graph
The effects on Boidhre's graph are like a smaller version of what bentw went through, because much of Boidhre's rating is tied to bentw right now.
Sometimes a widespread change in rank graphs is not because wms or the super admin changed the anchors,but because one or two anchors change. An anchor might stop playing or become weaker because of illness or something.
UnclMartin wrote:Sometimes a widespread change in rank graphs is not because wms or the super admin changed the anchors,but because one or two anchors change. An anchor might stop playing or become weaker because of illness or something.
Then why not use bots as anchors? WMS could buy a Zen or Crazystone or whatever version (or demand a free version if one wants to play it at KGS) and run them automatically on KGS.