We quantify the variability of faint unresolved optical sources using a catalog based on multiple SDSS imaging observations. The catalog covers SDSS Stripe 82, and contains 58 million photometric observations in the SDSS ugriz system for 1.4 million unresolved sources. In each photometric bandpass we compute various low-order lightcurve statistics and use them to select and study variable sources. We find that 2% of unresolved optical sources brighter than g=20.5 appear variable at the 0.05 mag level (rms) simultaneously in the g and r bands. The majority (2/3) of these variable sources are low-redshift (<2) quasars, although they represent only 2% of all sources in the adopted flux-limited sample. We find that at least 90% of quasars are variable at the 0.03 mag level (rms) and confirm that variability is as good a method for finding low-redshift quasars as is the UV excess color selection (at high Galactic latitudes). We analyze the distribution of lightcurve skewness for quasars and find that is centered on zero. We find that about 1/4 of the variable stars are RR Lyrae stars, and that only 0.5% of stars from the main stellar locus are variable at the 0.05 mag level. The distribution of lightcurve skewness in the g-r vs. u-g color-color diagram on the main stellar locus is found to be bimodal (with one mode consistent with Algol-like behavior). Using over six hundred RR Lyrae stars, we demonstrate rich halo substructure out to distances of 100 kpc. We extrapolate these results to expected performance by the Large Synoptic Survey Telescope and estimate that it will obtain well-sampled 2% accurate, multi-color lightcurves for ~2 million low-redshift quasars, and will discover at least 50 million variable stars.

41 pages, 14 figures, submitted to Astronomical Journal