An 18-year-old student in Philadelphia, Linus Chen-Plotkin, has used statistical analysis to demonstrate that Mozart's melodies are measurably less predictable, note-by-note, than those of Haydn, Beethoven and Schubert. The finding offers a rare data-driven answer to a question that has long occupied musicologists: what makes Mozart's music feel so distinctive and endlessly rewarding to hear, even after repeated listening. It doesn't replace traditional musicology but adds a quantitative dimension to largely subjective debates about the composers' relative genius.

Chen-Plotkin ran 600 movements from piano sonatas and string quartets by the four composers through bespoke software built on Markov chains and other statistical tools, finding that Mozart's tunes have a "shorter memory", making their progression harder to anticipate. He is cautious about overstating the results, noting they apply to small-scale melodic movement rather than structure across whole pieces, where Beethoven or Haydn might instead prove more unpredictable. Opposed to using generative AI in art-making, he sees the work as revealing patterns behind human creativity rather than reducing music to maths, and is starting university this autumn, having already begun a similar computational study of Chaucer.