Could it be that the technical nature of your website creates a biased sample? I'd estimate that the difference between Firefox and Chrome users in spoofing their user-agent and blocking trackers is far less than the bias created by sampling HN for example.