Imagine a quiet room where you need to pick out a single melody among many barely audible rustles. That's roughly what it's like to search for signals from extreme systems of two black holes — one huge, the other millions of times smaller. As these pairs spiral together, they send us gravitational waves, but their song lasts for years, and the notes are distorted by cosmic distances.

The main difficulty is that the mathematical "probability map" of such a signal resembles a mountain range with one sharp peak and many smaller ones. Standard methods often get stuck on false peaks, mistaking them for the real signal. The research team showed that these decoy peaks are not random: they cluster around the true target, like false paths in a maze that still lead toward the exit.

Using this pattern, the authors built a "smart" search in fourteen dimensions: that's how many parameters are needed to describe the system. First, they scan the whole area roughly, then, guided by the clusters of false peaks, they narrow the circle step by step. The method was tested on simulated data for the future space detector LISA. In noise resembling the real thing, the algorithm found two six-month signals, determining their position on the sky to within a tenth of a radian, and the distance with an error of about three percent. Thus, false clues have become reliable guides for the first time.