Sports Analytics Weekly by kubeia.io - 12/2025

John Oliver discusses sports betting, how it became so popular, why it can be so damaging, and – of course – how toned John’s shoulders are. Very toned. According to him.

In the era of positionless basketball, the traditional five-position model is becoming obsolete. The NBA has already begun reclassifying players based on skill sets rather than outdated positional labels, and, inspired by the work of Samuel Kalman, and Jonathan Bosch, we wanted to see if the same evolution is unfolding in the EuroLeague. Could a data-driven approach reveal new player positions that better define what actually leads to success? And beyond individual players, do certain roster structures consistently outperform others?

Traditional backtesting assumes that optimising a strategy on historical data and validating it on an out-of-sample period ensures future reliability. Traders typically backtest on in-sample data, optimise parameters, and validate on a brief out-of-sample period. If results look good, they assume robustness and move to live trading.

Machine learning is part of our products, processes, and research. But computers usually don’t explain their predictions, which can cause many problems, ranging from trust issues to undetected bugs. This book is about making machine learning models and their decisions interpretable.

Time series forecasting has undergone an extensive transformation from deterministic point predictions to sophisticated probabilistic approaches. This paradigm shift represents a fundamental change in how we deal with uncertainty in predictive analytics. Rather than generating single-point forecasts, modern probabilistic methods provide complete probability distributions of possible outcomes, thus enabling more nuanced and reliable decision-making processes.