I am trying to determine how many samples I would need in order to get 90% of all the subjects within +/-15% of a value. In other words, how many samples would I need to consider in order for a calculation (such as average) to be within +/-15% of their actual value?

In my case, I am trying to estimate the market value of a house based on other houses in the vicinity. What is the math behind determining the number of samples needed of nearby property values in order to extrapolate the market value of the house in question within a +/-15% level of accuracy? I know there are complicating factors behind real estate but I am looking at a reduced number of assumptions.

I seem to remember that was near the magic number for statistical significance although I haven't done this for 15+ yrs and I've forgotten how to do this properly!

Thank you.