Had an interesting email exchange with Tim Bray, who apparently believes that extrapolating from his weblog’s referrer logs is a reasonable way to estimate search engine market share (i.e. percentage of total search queries handled).
(Special prizes for the first readers to identify any of five ways in which this estimate could diverge from reality.)
Here’s a study with a larger, more diverse sample, though admittedly it lacks the transparency of Tim’s work. (Tim was kind enough to actually post his log analysis script.)