Sports Analytics Weekly by kubeia.io - 43/2024

Today, we publish the Lancet Public Health Commission on gambling—an inquiry and response to a neglected, understudied, and expanding public health threat. Gambling is not a simple leisure activity; it is a health-harming addictive behaviour. The harms associated with gambling are wide-ranging, not only affecting an individual's health and wellbeing, but also their wealth and relationships, affecting families and communities with potential lifelong consequences, and deepening health and societal inequalities. By assessing the barriers to preventing gambling-related health harms, the Commission unveils and deciphers the intersections between the social, commercial, legal, and political determinants of health.

Real Madrid enter El Clásico with uncertainty hanging over them. How do they replace Toni Kroos and also get the best out of Kylian Mbappé? Can Carlo Ancelotti re-establish balance in the team, or will their weaknesses bring their golden era to an end?

The 19th annual MIT Sloan Sports Analytics Conference will take place March 7-8, 2025 at the Hynes Convention Center.

Abstract:Understanding trajectories in multi-agent scenarios requires addressing various tasks, including predicting future movements, imputing missing observations, inferring the status of unseen agents, and classifying different global states. Traditional data-driven approaches often handle these tasks separately with specialized models. We introduce TranSPORTmer, a unified transformer-based framework capable of addressing all these tasks, showcasing its application to the intricate dynamics of multi-agent sports scenarios like soccer and basketball. Using Set Attention Blocks, TranSPORTmer effectively captures temporal dynamics and social interactions in an equivariant manner. The model's tasks are guided by an input mask that conceals missing or yet-to-be-predicted observations. Additionally, we introduce a CLS extra agent to classify states along soccer trajectories, including passes, possessions, uncontrolled states, and out-of-play intervals, contributing to an enhancement in modeling trajectories. Evaluations on soccer and basketball datasets show that TranSPORTmer outperforms state-of-the-art task-specific models in player forecasting, player forecasting-imputation, ball inference, and ball imputation. this https URL

Abstract:The Elo rating system is a popular and widely adopted method for measuring the relative skills of players or teams in various sports and competitions. It assigns players numerical ratings and dynamically updates them based on game results and a model parameter. Assuming random games, this leads to a Markov chain for the evolution of the ratings of the NN players in the league. Despite its widespread use, little is known about the large-time behaviour of this process. Aiming to fill this gap, in this article we prove that the process has a unique equilibrium to which it converges in an almost-sure sense and in Wasserstein metrics. Moreover, we show important properties of the stationary distribution, such as finiteness of an exponential moment, full support, and quantitative convergence to the players' true skills as the update parameter decreases. We also provide Monte Carlo simulations that illustrate some of these properties and offer new insights.

Abstract:In this work, three fundamentally different machine learning models are combined to create a new, joint model for forecasting the UEFA EURO 2024. Therefore, a generalized linear model, a random forest model, and a extreme gradient boosting model are used to predict the number of goals a team scores in a match. The three models are trained on the match results of the UEFA EUROs 2004-2020, with additional covariates characterizing the teams for each tournament as well as three enhanced variables derived from different ranking methods for football teams. The first enhanced variable is based on historic match data from national teams, the second is based on the bookmakers' tournament winning odds of all participating teams, and the third is based on historic match data of individual players both for club and international matches, resulting in player ratings. Then, based on current covariate information of the participating teams, the final trained model is used to predict the UEFA EURO 2024. For this purpose, the tournament is simulated 100.000 times, based on the estimated expected number of goals for all possible matches, from which probabilities across the different tournament stages are derived. Our combined model identifies France as the clear favourite with a winning probability of 19.2%, followed by England (16.7%) and host Germany (13.7%).

Abstract:Precise action localization in untrimmed video is vital for fields such as professional sports and minimally invasive surgery, where the delineation of particular motions in recordings can dramatically enhance analysis. But in many cases, large scale datasets with video-label pairs for localization are unavailable, limiting the opportunity to fine-tune video-understanding models. Recent developments in large vision-language models (LVLM) address this need with impressive zero-shot capabilities in a variety of video understanding tasks. However, the adaptation of image-based LVLMs, with their powerful visual question answering capabilities, to action localization in long-form video is still relatively unexplored. To this end, we introduce a true ZEro-shot Action Localization method (ZEAL). Specifically, we leverage the built-in action knowledge of a large language model (LLM) to inflate actions into highly-detailed descriptions of the archetypal start and end of the action. These descriptions serve as queries to LVLM for generating frame-level confidence scores which can be aggregated to produce localization outputs. The simplicity and flexibility of our method lends it amenable to more capable LVLMs as they are developed, and we demonstrate remarkable results in zero-shot action localization on a challenging benchmark, without any training.

Backtesting stands as a cornerstone technique in the development of systematic investment strategies, but its successful use is often compromised by methodological pitfalls and common biases. These shortcomings can lead to false discoveries and strategies that fail to perform out of sample. This article provides practitioners with guidance on adopting more reliable backtesting techniques by reviewing the three principal types of backtests (walk-forward testing, the resampling method, and Monte Carlo simulations), detailing their unique challenges and benefits. Additionally, it discusses methods to enhance the quality of simulations and presents approaches to Sharpe ratio calculations that mitigate the negative consequences of running multiple trials. Thus, it aims to equip practitioners with the necessary tools to generate more accurate and dependable investment strategies.

Decades ago, a mistaken assumption regarding statistical expectations sidetracked stock market analysts into applying casino-like statistical attributes to forecasting investor returns. This paper remedies that erroneous premise, introducing the Variable Expected Returns (VER) forecast, a closed-form distributionally-robust methodology that exhibits varying per-period expected returns. The VER methodology identifies the expectation of the geometric average of gross returns as the appropriate measure of central tendency for investors, in contrast to the flawed choice of the first raw moment in the classic compounded arithmetic return methodology. A novel series of calculations using S

In homage to John Hopfield and Geoffrey Hilton, Nobel Prize winners for their “fundamental discoveries and inventions that made machine learning and artificial neural networks possible,” I propose to explore the foundations of connectionist AI. We will examine how, by drawing inspiration from the functioning of the human brain, we can design algorithms capable of learning from data.

When data arise from a situation that can be modeled as a collection of n independent Bernoulli trials with success probability p, a simple rule of thumb predicts the approximate length that the longest run of successes will have, often with remarkable accuracy. The distribution of this longest run is well approximated by an extreme value distribution. In some cases we can practically guarantee the length that the longest run will have. Applications to coin and die tossing, roulette, state lotteries and the digits of π are given.