Can Computers Predict a Hit Song?
Critics of pop music frequently complain that the top hits are formulaic. But even if the song is predictable, is its success? Scientists from the University of Bristol’s Intelligent Systems Laboratory think so. They developed an equation that can predict whether or not a song will be a “hit.”
Using 50 years of data from the U.K.'s list of Top 40 singles, the scientists built a formula that they call the “hit potential equation.” It can predict, with 60-percent accuracy, whether or not a popular song will make it into the top 5, or if it will remain in the bottom 30 to 40 spots on the hit chart.
DNEWS VIDEO: IS IT FUTURE YET?
This is not simple science. The researchers use 23 measures, including tempo, loudness, harmonic simplicity, and “danceability.” These factors are then weighted for importance and summed to compute the hit potential for each specific song.
Because peoples' tastes in music changes over time — Hits today don’t sound like hits did in the 1980s — the scientists made their innovative method time-sensitive.
The calculations are run by computer algorithms designed to learn what factors are important to music trends at different times, and that is added to weight the predictive factors.
The equation performs the best in the first half of the 1990s and the past 10 years. However, in the later 1970s and early 1980s, it has trouble predicting hits, suggesting that music from that era had more variety and creativity. Danceability became an important factor for hits starting in the mid 1980s. Popular songs also tended to be louder than non-popular, but volume increased steadily until 2008, and has just recently started to decline.
The researchers continue to test their methods by making live predictions every week before the U.K. Top 40 chart is released. You can check their results here.
Photo Credit: Nicholas Monu/Getty Images