Relevant anecdotal evidence: I have a cousin who was really in to astrology a few years ago: so obviously my sister and I insisted she partake in an experiment. We had her do three specific readings (not just with signs but with the mercury rising nonsense for which she needed exact birth-dates and birth locations): for me, my sister and my brother who wasn't there. She read them to us without indicating who they belonged to and we tried to see if we could tell which ones referred to us. The second one she read was just shocking to hear. It described me perfectly. I was in awe for about 10 minutes until the experiment finished and I learned that the reading that described me perfectly belonged to my sister.

My girlfriend have been casually collecting data on this over the past 2 years. We occasionally end up in social situations with people who are flaky enough to take this stuff seriously. They'll usually -- after about 15 minutes of conversation -- make note of our 'auras' or personalities, and then guess a sign for each of us. We encourage them to try. Of the eight guesses we've had about our signs so far, none have been correct within 2 tries. I hope a few more years of this (and maybe some more data from less flaky friends) will offer enough data points to see if there's any bias, or if the odds of a good initial guess are uniform.