arohan - 2:15 am on Dec 31, 2012 (gmt 0)

Ah, both these reasons make perfect sense to me. I think the Chrome usage is statistically significant enough to generate this traffic data for them, even in cases when the sites do not have GA installed, specifically for high traffic sites. Where it might fail is for new, low quality and scraper sites with little traffic and Google seems to have difficulty with them lately.

I know Google technically claims that they do not use Chrome/Analytics data in their main algo, but they could be running this once in a while when enough data is collected (say once a year) and apply modifiers to their main algo.

Your 2nd reason, if true, throws up an interesting scenario. If many people visit a site, even when it does not rank well, Google might take it as a feedback that the site is good. When the rank improves, Google's share of traffic will increase again. I assume (hope) they will have some sort of a safeguard that prevents the Penguin "test" from running on the same site again. Again, hypothetical, but this would imply that changing domains to escape Penguin might be a losing proposition as it keeps resetting the clock, even if you start with a fresh link profile.

Perhaps these links that I am worried about might even be a positive in the next Penguin run as they DO get relevant traffic that converts. At least I can hope!