Speed Tests and Broadband Performance

In a light cap to a light-posting week, I wanted to note a post from Fiberevolution regarding speed tests. Many of us are frustrated with the available speeds and just how they compare to what we are promised. We have long seen a variety of criticisms of speed tests (including that operators like Comcast game the system by allowing greater speeds than one normally achieves to sites in practice), but few realized that one of the most well known tests (provided by Ookla) suffers from what appears to be a structural deficiency:
"Samples are sorted by speed, and the fastest half is averaged to eliminate anomalies and determine the result."
I wish my grades had been calculated like that back in school. I have come around to believing we need to collect greater data about what broadband we have available in order to compare providers (naturally, I suspect we will find that community networks do a better job of offering faster speeds). M-Labs has a good start and will be essential in this endeavor. More to come...