True, but the Goko leaderboard fluctuates a lot more, no?
A little more; I wouldn't say a lot. One way to interpret this is that Goko's leaderboard better reflects how you have been playing recently. I think the Isotropic leaderboard was too rigid, not allowing people to improve fast enough and allowed people to more easily "protect" high ratings through periods of bad play (not just bad luck) by only playing (other) highly ranked players.
I go through periods where I play badly (e.g. two weeks where I only play while watching a good tv show at the same time). My Goko rating drops a lot, but seems to be more accurate reflection of the skill I am actually displaying at the time.
Both Goko and Isotropish seem to function very similarly -- the differences seem to be because of differences in parameters rather than major differences between systems. They've told us in the past that at the heart the ranking is a mean minus some multiple of a standard deviation and it's updated in a Bayesian fashion using only win/loss/tie as the input. These things are true of TrueSkill, too.
And the Goko/MakingFun rep is completely right when he says wonky things happen with rankings in TrueSkill, too. You just don't see them presented to you on a game-by-game basis and much of the wonkishness is rounded off by presenting fewer digits. You effectively see 2 digits more information with Goko than Isotropish. Level 42. Level 5834. Maybe if they want to stem complaints, they should just change this.
WW has made some good arguments in other threads about just displaying and using the mean, too, for regular players, but that argument applies to many popular rating systems in use.
Of course, Goko is not taking advantage of the primary reason of having a rating system -- matching people of like skills to ensure more competitive games. That's what we should be pushing them on, especially since it was in the works at some point.