Sports Analytics Weekly by kubeia.io - 37/2025
- Week 37
- Week 36
- Week 35
- Week 33
- Week 32
- Week 31
- Week 30
- Week 29
- Week 28
- Week 27
- Week 26
- Week 25
- Week 24
- Week 23
- Week 22
- Week 21
- Week 19
- Week 17
- Week 16
- Week 14
- Week 13
- Week 12
- Week 11
- Week 10
- Week 9
- Week 7
- Week 6
- Week 5
- Week 4
- Week 3
- Week 2
- Week 1
đź“ť Sports Analytics
Expected goals and the Monte Carlo Trap
kuleuven.be - Jesse Davis, Jan Van Haaren, Pieter Robberechts
On a team level, xG is often used to assess the quality of the goal scoring opportunities that a team creates and concedes. That is, it helps assess performance independent of whether or not a scoring opportunity resulted in a goal. In this regard, there are two interrelated canonical use cases: (1) computing post-hoc win-draw-loss probabilities based on the teams’ chance quality, and (2) constructing expected points (xPoints) league tables based on these probabilities.An obvious way to compute the win-draw-loss probabilities is to approximate them using a Markov Chain Monte Carlo (MCMC) simulation (c.f.: Danny Page’s Match Expected Goals Simulator). However, as Jonas Lindstrøm, Tony ElHabr and others have pointed out, it is actually possible to compute them exactly using a Poisson Binomial distribution.This blogpost has two goals:Part 1 compares the relative performance of the MCMC and exact approaches in terms of accuracy, stability, and concrete use cases.Part 2 compares the runtime efficiency of various ways of implementing the MCMC approach.
🤖 Machine Learning
Justify your answer
argmin.net - Ben Recht
Now, in machine learning, we almost never have explicit probabilistic models of data and outcomes. In machine learning, prediction is a missing data problem. There is a population of outcomes on which we’ll be tested, and we want to find a prediction function that has low error when averaged over this population. As we saw last week, the optimal prediction rule still retains the same form as above. It just has a very different interpretation.
🎙️ Podcast
Pinnacle Odds Dropper Review
smartbettingclub.com - Peter Ling
In the latest SBC Podcast episode, I sit down again with Dylan, co-founder of Pinnacle Odds Dropper (POD), to take a deep dive into sharp betting, market efficiency, and the lessons from analysing over 345,000 logged bets.We explore how the software helps advantage bettors track steam moves, why Pinnacle remains the barometer for bookmakers, and what our August 2025 independent SBC review uncovered. Dylan also shares insights on CLV, market limits, niche sports, and the growing challenges facing US bettors under new tax rules.
🧮 Statistics
Your noise is my signal
argmin.net - Ben Recht
My favorite example to motivate actuarial prediction is sports statistics. In particular, solo statistics like free-throw shooting. For a particular player, predicting the likelihood they’ll make a free throw can be well estimated from their history of free throws and not much else. Your confidence that a player will make a shot is based on how frequently they have made that shot before.
New statistical tool enhances prediction accuracy
phys.org - Stephanie Baum
An international team of mathematicians, led by Lehigh University statistician Taeho Kim, has introduced an innovative method that could significantly improve how scientists make predictions, especially in fields like health, biology, and the social sciences.The new approach is designed to make predictions that better agree with actual outcomes. Based on this idea, researchers named it the Maximum Agreement Linear Predictor, or MALP. This prediction approach achieves higher agreement in predictions by optimizing the concordance correlation coefficient (CCC), which measures how well pairs of observations fall on the 45-degree line of a scatter plot, combining both precision—how tightly points cluster—and accuracy—how close they are to the line.