Identifying Banksy using statistics

What the article doesn't clarify is that the researchers mapped the locations of Banksy pieces against hand-picked locations of interest, such as the home addresses of people who were in the past said to be Banksy.

So, the fact that the researcher's model "fingered" Robin Gunningham is not as surprising as the Economist makes it. In fact, Gunningham was "fingered" by the model because he had been fingered in the past by others.

The Economist's article fudges this considerably and makes it seem like the model can just converge on a correct location on a map just from the signal, which is very doubtful.

At best this is a good guess as to what past guess is a good guess; in this case, about Banksy's identity. In the future, about the usual suspects, probably.

Edit: Original study (behind a paywall): http://www.tandfonline.com/doi/abs/10.1080/14498596.2016.113...

Here, we use geographic profiling, a statistical technique originally developed to prioritise large lists of suspects in cases of serial crime such as murder, rape and arson, to assess the evidence supporting one prominent candidate.

The research identified 10 "suspect sites" that were used as _input_ to the model. Most of them are associated with Gunningham:

Suspect sites in Bristol included Gunningham’s house in the Easton area of the city, The Plough in Easton (for whom Gunningham played football), and their playing fields at Baptist Mills Primary School, as well as Gunningham’s old school, Bristol Cathedral School.