The DNA tests are done for what are known as "Short Tandem Repeats". These are spots in the DNA where the same sequence (2-6 base pairs) is repeated several times. Unrelated people have different numbers of repeats.

They're looking at a dozen or more locations. The number of repeats at any location is unrelated to the number of repeats at another location.

So, if you've got a number between 2 and 10 or 20, you've got 10% to 5% of the population that has that particular number of repeats. 1/(0.1 ^ 13) is 10 trillion which is the ballpark for the number of variations for STRs possible on a 13 location test.

The data that is stored looks like [6 4 12 16 2 15 19 18 14 5 3 7 8]

https://en.wikipedia.org/wiki/Microsatellite

https://en.wikipedia.org/wiki/STR_analysis

Well, that's not how its really used per say. They aren't looking for behavioral features, they are looking for pattern matches not unlike a fingerprint. The advantage of this versus a fingerprint is the heritability aspect. You share half your DNA with either parent and you can proportionally estimate degree of shared dna for relatives in a geneology. That means they don't have to make sure the killer is in the finger print database, a distant relative and some math can in practice expand your database.

> They aren’t looking for behavioral features.

I might be on board if you could make the stronger claim that they can’t look for behavioral features nor will they be able to in the future. A good hash provides that guarantee.

The argument is that they aren't considering phenotype information when they are solving crimes. How would that information even be used? A lawyer would have a stronger argument arguing a predisposition to criminality on environmental factors: a broken home, being raised when leaded gas was used, etc, versus arguing in terms of genetic association. These studies are not very robust connecting behavior and genetics because environment is a huge factor that explains the observed variances in these phenotypes.

That being said, these data are being used for actually legally actionable things: first order pattern recognition using the same legal frameworks as fingerprints. Not second order "what do these patterns actually do in the body" or third order "what do these functions in the body do for the individual", each of which would probably take a PhD project to work out.