From The Vault: Machine Learning in Elite Sports
In this presentation, we hear Professor Sam Robertson discuss how machine learning principles can be used to inform the way we should consider and address problems in sport.
At the time Sam was the Head of Research and Innovation at Victoria University and the Western Bulldogs Football Club. Throughout the talk he draws upon many applied examples of how machine learning is currently advancing elite sport settings, by allowing them to identify the optimal course of action on various problems, improve their organisational decision making, and decrease the factor of human error.
Throughout his presentation he discusses four main topics:
1. Confusing domain expertise with decision-making expertise [00:07:39]
We shouldn’t expect domain experts to also be decision-making experts: in fact, doing so is probably harmful to your organisation.
Machine learning isn’t perfect, but in general it will perform better than other forms of analysis.
2. Framing the question appropriately [00:12:28]
Remember to ‘question the question’. Consider machine learning as a philosophy for problem framing.
Always assess the operational compatibility of both the question and the solution. Consider the format in which data is represented.
Develop an evaluation framework to obtain a comprehensive understanding of the solution’s efficacy.
3. Bounded rationality [00:24:37]
The more good quality data, the better. But the only way we can handle and interpret this data is via machines.
All models are biased, as we will never include all of the information that we need to consider.
It’s not about being perfect; its about being better.
4. Interpreting complexity [00:34:54]
Comparing human vs machine based decision-making is like comparing apples and oranges - they utilise different informations sets.
Whilst we should embrace complexity, we should also strive for parsimony.
Interested in the material referred to in this presentation?