You hear a new song. Will it be a hit or a flop? Researchers from Bristol University in the U.K. say they can now tell you - well, sort of. After studying the Top 40 singles charts over the last 50 years and examining the audio characteristics for hits and flops, the team has come up with a formula as to what makes for a successful song and used it to devise software that "predicts" hits. The next step is a web app to allow budding musicians to score their own songs.

The prediction characteristics include musical features such as time signature, tempo, beat-variation, "danceability," as well as the harmonic simplicity of the songs and how noisy/loud they are. Importantly, these variables are examined for how they shift over time so in a sense, it is a shifting formula.

As Dr. De Bie, Senior Lecturer in Artificial Intelligence at Bristol University explains, "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." For example, the ability to dance to a song was a huge determining factor of how well they would fare on the charts in the 1980s, but this characteristic became less and less important as music moved into the early 90s rock ballads.

In the 80s, when slower musical styles (tempo 70-89 beats per minute) were more likely to become a hit, Simply Red's version of If You Don't Know Me By Now, with its mix of slow tempo, harmonic simplicity, and sing-along chorus ticked many of the boxes for success - even before it was released.

The team's website, scoreahit.com, explains that their prediction system is based on regression: "mathematically the hit potential (peak UK chart position) of a song is denoted by a variable (y) and a set of audio features (x) of the song are also presented. A pre-trained classifier f(x)=w'x is then used to estimate the hit potential." Got all that? If not, the website provides an even more complex mathematical table to confuse you further.

However, the bottom line (as outlined in a paper presented last week to an international workshop) is the system's prediction accuracy of 60 percent. That's a long way from perfect, but better than chance and as the website says, "We know that we're never going to be able to predict every song which peaks in the top 5."

A detailed graph of the hit predicting system running from the 1960's until 2010

Regardless, the formula seems to have the highest success rate of known attempts to predict chart toppers. No doubt, that is in no small part due to it taking account of the way the musical-landscape shifts over time. How these musical features move over time can be seen in the embedded video

The Scoreahit.com website provides an interesting look back in time, not only showing songs that from their audio characteristics were destined to chart well, but also those that went against the grain. Songs that reached the top 5 despite going against the formula included such chart-toppers as Luciano Pavarotti's Nessun Dorma, and Guns'N'Roses' November Rain.

A major problem with this system for budding songwriters is that (so far) it can only measure the characteristics of a fully constructed piece of music.

If audio characteristics such as loudness are to be properly evaluated, then it appears that the song would need to be fully written, produced, mixed and mastered before it could be properly assessed - with the consequent expenditure of time, money and effort that entails.

The Bristol University method is just the latest amongst many methods attempting this vexed question including:

• Hit Song Science: Another statistical technique that looks at trends, styles and sounds. At this website you can upload a song and (for a price) get a score and a report
• Music Xray: Using what appears to be a form of Cluster Analysis, this website is reviewed in this blog

It should be noted that not all of the attempts at predicting hits focus on deconstructing the DNA of the song's characteristics. There are other approaches that take a completely different tack such as, which relies on linguistic analysis of online reviews of the song. Perhaps most impressive of all however, is the Emory university group approach, which uses MRI scans of people's brains while they are listening to the song as the basis for the predictions. As impressive that may sound technologically, it is however a prohibitively expensive way to go about predicting hits ... unless of course you're Simon Cowell.

By contrast, the Bristol University researchers now say they are working on an economical app to allow budding musicians to score their own songs using the Scoreahit formula.