Several people I know have said Astrology is surprisingly accurate.

You can't really depend on recommendations for this sort of thing, since there is a force (vagueness of the reports) which explains away the evidence of the recommendations.

The Myers-Briggs Type Indicator is a popular but flawed way of understanding your personality.

I use my results as a checklist when I'm disagreeing with someone to see why we might be seeing things differently. It seems to help, although I haven't tested rigorously.

What did the mean with accuracy? The test told them accurately what they thought of themselves? To the extend that a test tells you something new that derivative from what you think of yourself the test should seem inaccurate.

I've tried it and I'm not sure what surprisingly accurate would mean as the questions were precisely of the sort you knew what they were asking for. It seems it could be useful for finding like minded people with the same four letter label though.

From what I've heard, the Myers-Briggs test is fairly accurate, but even the paid MBTI evangelists say that reading through detailed descriptions of the various types to find the one which feels most familiar is better. The detailed description part is important, because the details (about a page per type, IIRC) go into both the strong and weak points, providing pressure to reject descriptions which have flaws that feel inaccurate.

From what I remember, the corporate paid evangelist visits generally do type testing three ways: Written test, description-based given by your peers, and description-based self-report. The first is for calibration, the third is the main assigner, and the second is basically a conversation-starter; the idea is to discuss ways in which people feel different internally from how they're perceived.