Divide by n - 1, where n is the number of data points. A long time ago, statisticians just divided by n when calculating the variance of the sample. This gives you the average value of the squared deviation, which is a perfect match for the variance of that sample.
