Equation reveals hit song potential
Published 17/12/2011 | 00:12
A computer programme could soon be challenging music tycoon Simon Cowell's "hitman" reputation.
Scientists at the University of Bristol have come up with a "hit potential equation" that can predict a top five hit with 60% accuracy.
The formula brings together musical features such as tempo, timing, song duration and loudness, as well as more detailed information such as chord and harmonic complexity.
Run through a computer, it produces a score that indicates whether a song will reach the top of the UK singles charts or never rise above position 30 or 40.
The researchers used artificial intelligence software to analyse 50 years of top 40 singles, looking for the features that distinguish a "hit" from a "miss".
The study generated a list of "weights" (w) attaching levels of importance to 23 song elements. These were applied to features (f) of the new song being tested to produce the formula: Score equals (w1 times f1) plus (w2 times f2) plus .. plus (w23 times f23)
The research, presented at an international workshop on Machine Learning and Music in Sierra Nevada, Spain, uncovered traits that helped songs become hits in previous decades.
Team leader Dr Tijl De Bie, senior lecturer in artificial intelligence at Bristol University, said: "Musical tastes evolve, which means our 'hit potential equation' needs to evolve as well. Indeed, we have found the hit potential of a song depends on the era. This may be due to the varying dominant music style, culture and environment."
The scientists found that before the 1980s, song danceability was not very relevant to hit potential. From the 1980s onwards, danceable songs were more likely to become hits.
Slower styles, such as ballads, were also more likely to be hits in the 1980s. From the 1990s onward, hits tended to have simpler, binary rhythms such as 4/4 time.