http://en.wikipedia.org/wiki/Dna_fingerprinting#STR_analysis
The true power of STR analysis is in its statistical power of discrimination. In the US, there are 13 core loci (DNA locations) that are currently used for discrimination in CODIS. Because these loci are independently assorted (having a certain number of repeats at one locus doesn't change the likelihood of having any number of repeats at any other locus), the product rule for probabilities can be applied. This means that if someone has the DNA type of ABC, where the three loci were independent, we can say that the probability of having that DNA type is the probability of having type A times the probability of having type B times the probability of having type C. This has resulted in the ability to generate match probabilities of 1 in a quintillion (1 with 18 zeros after it) or more. However, since there are about 12 millions monozygotic twins on Earth, that theoretical probablitity is useless. For example, the actual probability that 2 random persons have the same DNA is only 1 in 3 trillions.