An Irish data scientist has developed a new artificial intelligence based system that can use runner’s race data to make predictions that can help them improve their personal best times.
A marathon runner himself, Professor Barry Smyth from the Insight Centre for Data Analytics has been collecting and analyzing race data from millions of marathon runners around the world for some time.
This has led to insights about race behavior, including that women are more disciplined than men when it comes to pacing themselves and that the speed gap between the genders closes with age.
He also discovered that those who sprint to the finish tend to end the race more slowly overall.
Professor Smyth then applied artificial intelligence techniques to what he had learned to see if he could predict how runners could improve their personal best times.
“To be clear, we were not talking about predicting any old marathon finish-time, based on a recent half-marathon or 10k race time; there are plenty of race calculators available to do this,” he said.
“Rather, we were interested in predicting a challenging but achievable PB time, which a runner might be capable of pushing themselves to achieve.”
This involved predicting a challenging but realistic finish time as well as coming up with a pacing plan to achieve it, all tailored to the specific course being run.
He developed a model that used Case-Based Reasoning – a form of machine learning that adapts times and pacing profiles of runners who are similar to those he is trying to predict a better personal best for.
The system takes data from the past races of the benchmark runners who used to run similar times to those who Professor Smyth is trying to help, but who went on to improve their speeds.
Professor Smyth applied his model to data from the London marathon and found those who ran it in 150 minutes could improve their time by about five minutes.
Among men who completed it in four hours, a 22 minutes improvement was possible, with 17 minutes being the new personal best for women running at the same pace.
The system improves the more data from different runners and courses that is added to it and Professor Smyth is working on improving its accuracy and other applications for the model.
“We are looking at many different ways in which machine learning can be used to help marathon runners and other athletes, not just for personal best prediction, but also for injury prevention, for recovery advice, to personalize training plans and so-on,” he said.