Rate Clustering

[Brian the Roman]

One of the things I expected to see in the stats was a clustering of people all doing the same speed. Stands to reason that over time, you'll pass all the people going slower and fall behind the faster people. So one would expect to eventually reach a stable position where one rarely passes anyone and is rarely passed by someone else. I don't see any evidence of this in the stats at all. Clearly, varying speeds of an individual over time wil screw this up as will speedy latecomers and dropouts. But I'd expect to see some clustering... Maybe we haven't been at it long enough. Has anyone else seen any evidence of this in other longer-running projects, like SETI?

ms

[IronBits]

Everyone jumps around to several projects, depending on mood, gauntlets, team rallies etc.
It's never a constant... then there is the 'ramp ups', where folks go out and buy a bunch more boxen because their neighbor is getting too close. :D Only to find out the cost of electricity to run that many more, so they sell them off :(
Then the Servers go down, stats get borked, or computer crash because they tweaked them to much...
Endless variables... most of the power pharms are at the top tho ;)

[Kileran]

i ran a constant rate on distributed net's RC5 64 for over 2 years, i never reached a stable position.

with thousands of people contributing at odd rates, new people comming everyday, some people leaving, the stats will never settle down.

[TheOtherZaphod]

I am pretty close to equilibrium. I am in 10th place overall, and on statman and Dyyryaths graphs my production rarely strays more than a few positions from there on a day to day basis. But, there are four people in front of me who I am slowly catching up to, and an equal number behind who are catching up to me.

It isn't an inherent mathematical function that keeps things from stabilizing, it is people; you just can't count on them.

It is also true that this project is pretty young, but that the higher up you go; the longer you hold any given position. That, I believe, does involve a purely mathematical function.

If we all started over again, you would see significant grouping for the first few weeks, but after that things would start to randomize again.

Just my .02.