True, but I'm not a scientist, I'm certainly not claiming that this was done to be scientific (it was done in five minutes, while reading an HN thread), and this isn't worth doing much more than that. If I care enough, I know how to do it a lot more exhaustively. For example, I'd probably write my own speed test, and host it in a known place. Relying on known tests; even obscure ones, is not something that I'd suggest for someone interested in The Scientific Method. The results of that Ookla test suggest that there may some "emissions testing" stuff going on.