For millions of football fans across the UK, match day brings excitement and often a cheeky punt on the outcome. But in a time of a tricky job market and a willingness, by many, to retrain into a professional career, it’s difficult to see how betting could play any kind of role.
Well, it can, but not in the way you’re thinking. Forget being a professional gambler. What we can do instead is use our hobby, and a data-centric pastime, as the basis for learning how to become a data scientist.
The perfect dataset
Football produces an abundance of structured data that’s ideal for beginners. The online betting apps themselves do a good job of providing live odds, but there are vast datasets we can tap into. Each match has dozens of measurable variables: goals scored, shots on target, possession percentages, corner kicks, yellow cards, xG, and so on. Historical league tables, transfer values, and injury reports add more layers.
This wealth of information isn’t just a great place to source data (any data), but it’s contextual. We already understand it, and with that context, any new skills we learn in how to manipulate, structure or analyse that data will be far more relevant to us. It’ll sink in better, and we won’t forget it.
It’s also far easier to spot anomalies or mistakes in Manchester United’s defensive statistics than it is in abstract sales figures from an unfamiliar industry.
Statistical concepts through probabilities
Betting odds are the clear and immediate introduction to probability theory and statistical thinking. It’s a good way to first and foremost understand markets and the Wisdom of the Crowd, in which odds, with sufficient betting volumes, lead to an implied probability that is surprisingly accurate. This introduces the concept value and expected value, and whether you are more sure of an outcome than the average punter/market price is.
Predictive modelling
Of course, it’s all about predictions, and so too is data science, whether you’re predicting the next late delivery or predicting who will not pay their invoice on time. Football matches are ongoing, and so we can test predictive models on upcoming games, but also backtest our strategies.
You might start with simple linear regression, or whatever you’re first taught in a free online course, using home advantage and recent form to predict goal differences. More sophisticated approaches could incorporate Poisson distributions for goal scoring or logistic regression for win/draw/loss predictions. Of course, there is plenty of room for machine learning techniques too.
Real-world skills
The analytical skills developed through football betting analysis may just seem like a proxy or a means to an end, but it is an end in and of itself. The job market for data scientists within football clubs is booming, just as AI hype is spiralling chip prices, as they acquire data-driven approaches to training, injuries, opponent analysis, transfers, and much more.
Of course, the skills are transferable to other industries too. Data cleaning techniques learned while standardising player names across different databases apply to any Business dataset. The statistical methods used to identify value bets translate directly to marketing attribution, fraud detection, and customer segmentation challenges.
Football betting shouldn’t be your career – but it could be the perfect context that launches your journey into data science.