Predicting Eurovision Song Contest Results: A Hit Song Science Approach

Abstract

Prior studies on hit song prediction have predominantly focused on forecasting a song’s success in music charts, neglecting the examination of song contests such as Eurovision. This paper presents a framework for predicting Eurovision result rankings, with a particular focus on the semi‑finals, which determine qualification for the grand final, and the rankings of the grand final. By integrating intrinsic song characteristics, public appeal, and contest‑specific data, the study evaluates seven feature sets across multiple years of Eurovision, spanning from 2008 to 2024. The inclusion of features such as audio and lyrics attributes, YouTube daily views, the previous year’s vote ratio and vote reciprocation, and performance order provides a multi‑modal approach to understanding song success in the contest. Key findings indicate that the intrinsic song features employed in this study alone do not accurately predict rankings, as they account for only a minimal portion of the variance in contest results. While public appeal, represented by YouTube daily views, emerged as a significant factor, it may be influenced by post‑contest exposure bias. The most effective prediction model combined intrinsic song characteristics, public appeal, and contest‑specific data, yielding the most consistent results across semi‑finals and grand finals over multiple years.

Keywords

• eurovision song contest
• hit song science