What brought Professor Stephen Clarke to prominence were
his footy tips, generated by a computer program he created
in 1980. Its predictions, with a 68 per cent success rate, have
appeared continuously in newspapers or on TV ever since.
Photo by Michelle Templeton

In the late 1970s, Swinburne University of Technology’s Professor Stephen Clarke had yet to create the AFL tipping program that enlivens morning tea conversations in workplaces around Australia during the ‘footy’ season. But his passion for sports statistics was already well known to his students, and became the foundation of Swinburne’s expertise in the collection, analysis and provision of statistical sports information.

Back in those days, Professor Clarke was a keen squash player. As a statistician and a player he became interested in the two options the receiver can choose if the score reaches 8–all. Is it better for the receiver to choose to play to nine? Only the server can win points, so the game could be over in the next rally. Or does going an extra two points to 10 improve the chances of snatching a narrow victory?

Professor Clarke found that playing to 10 is usually better, but not if your opponent is thrashing you. If you are struggling to win even two rallies in five, it is better to choose the more ‘sudden death’ option of nine.

However, what brought Professor Clarke to prominence in the world of sports stats were his footy tips, generated by a computer program he created in 1980. Its predictions, with a 68 per cent success rate, have appeared continuously in newspapers or on TV ever since.

Professor Clarke wanted to excite his students by applying statistics to everyday life. He intentionally kept his program simple to avoid tedious hours at the keyboard, and to simplify the concepts. "All that the program has to be told each week are the final scores, and the grounds on which matches are played," he says.

His influential statistical analysis of home-game advantage helped establish an international reputation in sports stats. "In all team sports, playing at home improves the odds," he said. "So it had to be factored into the program."

The program was simple enough to run quickly, even on the sluggish computers of almost three decades ago. The vast processing power of today’s systems adds nothing to the program’s accuracy and only a little to its speed. Even the most computer-intensive part of the output – predicting the rankings of the final eight – takes only minutes, as the computer ‘plays’ out the rest of the season 10,000 times.

In most seasons, a few savvy media commentators pip Clarke’s program’s predictions, but those inspired (or are they just lucky?) individuals cannot keep it up, season after season. The computer is consistent year in, year out, even though its task has become more difficult. In the early 1980s, three AFL teams were regular contenders for the wooden spoon. They rarely won except against each other. Tipping their results was easy. Today the competition is closer, raising the bar for tipsters.

Swinburne’s expertise in sports stats now embraces rugby, cricket and tennis. Since 2001, Swinburne has marketed this intellectual property through Sportsbet 21, a service company which is part-owned by the university. It provides sports data modelling for the UK betting agency Ladbrokes under a licence agreement that delivers a revenue stream to the company.

The commercial value of Swinburne’s models to the betting industry is that they can be used to calculate odds on outcomes such as the number of runs in the next over in one day and 20/20 cricket, or the score in the next game of a tennis tournament. The odds required for this ‘micro-betting’ need to be available almost instantly, which rules out human calculations.

Swinburne’s know-how in sports data processing and interpretation was sought out by Ted Hopkins, founder and creative director of Champion Data, the official provider of statistics for the AFL.

Mr Hopkins played in Carlton’s 1970 premiership team and has studied and written about sports statistics for many years. In the early 1990s, he saw the potential of computer processing and Internet delivery for sports stats. With Professor Clarke consulting, Champion Data revolutionised the collection, processing and presentation of these statistics.

As each AFL match is played, another team, of eight people – some at the match, others in a company backroom – collect and monitor data. A minimum of 30 statistical events is recorded and other useful statistics are derived from the raw data. This information is instantly fed to the media, clubs and scoreboards over the Internet, and is studied by coaches as they refine strategies in the heat of the match.

Swinburne’s sports stats research program continues to attract PhD students, some supported by Champion Data. One of these projects investigated prediction of scoring events based on computer recognition of scoring patterns.

Another PhD student is analysing spatial information about the parts of the playing field where most activity occurs, creating maps of ‘hot spots’ – regions of intense play. This project may uncover links between team tactics and improved scoring, and take the science of coaching to new heights.