Sports Analytics Weekly by kubeia.io - 38/2023

Billy’s book, Gambler: Secrets from a Life at Risk, is his story, describing how a poor country boy from Kentucky evolved from a compulsive gambler and alcoholic to become a successful businessman, philanthropist, and sports betting legend.

I love reading stories about individuals, or groups of people, identifying opportunities in the sports betting markets. I wrote about the "Hole-in-One" gang almost 13 years ago, a post which also mentioned the mispricing by bookmakers of the 3:3 draw in football, both great examples of people spotting a pricing error and making money out of it.There was another story arguably in this category on Sunday in the NFL, with the sportsbook FanDuel offering 200-1 on every team in the 12 afternoon games to make at least one field goal.

The Spanish National Police, in cooperation with the Spanish Tax Agency, Europol and INTERPOL, have dismantled an organized crime group suspected of fixing sporting events as well as using technology to place bets ahead of bookmakers.So far, 23 suspects have been arrested, including one of the group’s leaders who was apprehended on the basis of an INTERPOL Red Notice for persons wanted internationally.The operation began in 2020 when Spanish officers detected a series of suspicious online sports bets placed on international table tennis events. After analysing available data, investigators identified a criminal network of Romanian and Bulgarian origin.Members of this crime ring fixed matches outside of Spain by corrupting athletes. Once the outcomes were agreed, crime group members based in Spain would then place online bets on a massive scale.

Perched forward in his chair, scanning the room as if to assess his persuasiveness, Mikel Arteta delivered a post-match explanation that will in future come to be viewed as one of two things: a visionary reframing of the idea that number one status belongs to a single goalkeeper, or a sticking plaster on a delicate situation too fragile to unpack.

Max Scherzer made a dark prediction at the beginning of the baseball season. Pitcher injuries, already at high levels, would intensify. His concern, Scherzer repeatedly warned to anybody who would listen, was the introduction of the pitch clock.  

One of the quickest ways to make an impact on a college football team is transfer portal recruiting. The ability to sign players who are older, more established, and more experienced shortens the time frame needed to turn a program around. The early returns on Colorado and Texas State’s huge transfer signing classes are shining examples of successful transfer portal signing classes.

Call​ it the Jaap Stam conundrum. The story is told in Rory Smith’s entertaining book about the use of data in football, Expected Goals. Stam was a very talented but ageing central defender who had for three years been crucial to the success of Alex Ferguson’s Manchester United. Stam got into Ferguson’s bad books, and Ferguson decided to get rid of him, backed by data showing that Stam was now making fewer tackles. Stam was sold to Lazio for £16.5 million. Simon Kuper, co-author of Soccernomics (2009), about the use of numbers in football, wrote that this might have been the first deal in football based partly on data. It was, as Ferguson later admitted, a ‘bad decision’. Stam went on to have several productive years at the top of Italian football. The mistake was based on the fact that the data were Delphic. The question Man U wanted to have answered was ‘Should we keep Stam?’ Instead, the oracle answered the question ‘How many tackles is Stam making?’

Abstract:Evaluating the performance of human is a common need across many applications, such as in engineering and sports. When evaluating human performance in completing complex and interactive tasks, the most common way is to use a metric having been proved efficient for that context, or to use subjective measurement techniques. However, this can be an error prone and unreliable process since static metrics cannot capture all the complex contexts associated with such tasks and biases exist in subjective measurement. The objective of our research is to create data-driven AI agents as computational benchmarks to evaluate human performance in solving difficult tasks involving multiple humans and contextual factors. We demonstrate this within the context of football performance analysis. We train a generative model based on Conditional Variational Recurrent Neural Network (VRNN) Model on a large player and ball tracking dataset. The trained model is used to imitate the interactions between two teams and predict the performance from each team. Then the trained Conditional VRNN Model is used as a benchmark to evaluate team performance. The experimental results on Premier League football dataset demonstrates the usefulness of our method to existing state-of-the-art static metric used in football analytics.

Abstract:The success of a football team depends on various individual skills and performances of the selected players as well as how cohesively they perform. We propose a two-stage process for selecting optimal playing eleven of a football team from its pool of available players. In the first stage a LASSO-induced modified multinomial logistic regression model is derived to analyse the probabilities of the three possible outcomes. The model considers strengths of the players in the team as well as those of the opponent, home advantage, and also the effects of individual players and player combinations beyond the recorded performances of these players. In the second stage, a GRASP-type meta-heuristic is implemented for the team selection which maximises its probability of winning. The work is illustrated with English Premier League data from 2008/09 to 2015/16. The application demonstrates that the model in the first stage furnishes valuable insights about the deciding factors for different teams whereas the optimisation steps can be effectively used to determine the best possible starting lineup under various circumstances. We propose a measure of efficiency in team selection by the team management and analyse the performance of the teams on this front.

Abstract:Player identification is a crucial component in vision-driven soccer analytics, enabling various downstream tasks such as player assessment, in-game analysis, and broadcast production. However, automatically detecting jersey numbers from player tracklets in videos presents challenges due to motion blur, low resolution, distortions, and occlusions. Existing methods, utilizing Spatial Transformer Networks, CNNs, and Vision Transformers, have shown success in image data but struggle with real-world video data, where jersey numbers are not visible in most of the frames. Hence, identifying frames that contain the jersey number is a key sub-problem to tackle. To address these issues, we propose a robust keyframe identification module that extracts frames containing essential high-level information about the jersey number. A spatio-temporal network is then employed to model spatial and temporal context and predict the probabilities of jersey numbers in the video. Additionally, we adopt a multi-task loss function to predict the probability distribution of each digit separately. Extensive evaluations on the SoccerNet dataset demonstrate that incorporating our proposed keyframe identification module results in a significant 37.81% and 37.70% increase in the accuracies of 2 different test sets with domain gaps. These results highlight the effectiveness and importance of our approach in tackling the challenges of automatic jersey number detection in sports videos.

Abstract:The long-standing problem of novel view synthesis has many applications, notably in sports broadcasting. Photorealistic novel view synthesis of soccer actions, in particular, is of enormous interest to the broadcast industry. Yet only a few industrial solutions have been proposed, and even fewer that achieve near-broadcast quality of the synthetic replays. Except for their setup of multiple static cameras around the playfield, the best proprietary systems disclose close to no information about their inner workings. Leveraging multiple static cameras for such a task indeed presents a challenge rarely tackled in the literature, for a lack of public datasets: the reconstruction of a large-scale, mostly static environment, with small, fast-moving elements. Recently, the emergence of neural radiance fields has induced stunning progress in many novel view synthesis applications, leveraging deep learning principles to produce photorealistic results in the most challenging settings. In this work, we investigate the feasibility of basing a solution to the task on dynamic NeRFs, i.e., neural models purposed to reconstruct general dynamic content. We compose synthetic soccer environments and conduct multiple experiments using them, identifying key components that help reconstruct soccer scenes with dynamic NeRFs. We show that, although this approach cannot fully meet the quality requirements for the target application, it suggests promising avenues toward a cost-efficient, automatic solution. We also make our work dataset and code publicly available, with the goal to encourage further efforts from the research community on the task of novel view synthesis for dynamic soccer scenes. For code, data, and video results, please see this https URL.

Volatility estimation and forecasting plays a crucial role in many areas of finance.For example, standard risk-based portfolio allocation methods (minimum variance, equal risk contributions, hierarchical risk parity…) critically depend on the ability to build accurate volatility forecasts1.Multiple methods for estimating volatility have been proposed over the past several decades, and in this blog post I will focus on range-based volatility estimators.

You can now write Python code to analyze data, build machine learning models, and create visualizations within Excel spreadsheets.

“Feature Importance” is not enough. You also need to look at “Error Contribution” if you want to know which features are beneficial for your model.

But it goes well beyond that, for over a decade scikit learn devs made incorrect and false claims that some models like “logistic regression returns well calibrated predictions by default as it directly optimises log loss.”

The general aim of the project is to develop a machine learning tool to help people to play against bookmakers and increase their chances of winning in sports betting. A machine learning model predicting the sport outcomes in National Hockey League (NHL) is going to be developed. The data set necessary for the model training is taking from the sports web sites using web scraping methods available in Python and Selenium. As compared to the available NHL data sets the new one contains not only the sport statistics, but also the bet statistics. This allows to estimate not only the accuracy of the model as compared to the known true outcomes, but also compare the efficiency of the developed model with models used by bookmakers. The impact of the matches history for a particular team (team’s form) on the performance of the model will be analyzed. The classification problem will be solved with random forest, gradient boosting and neural networks. The best model will be chosen to predict the sport outcomes. Unsupervised learning techniques (clustering) will be applied to the data set in order to explore the hidden patterns in the data in order to check the hypothesis about the increasing probability of draws for some particular distribution of points. A new extended data set will be formed containing the model predictions and bookmakers predictions. Reinforcement learning (Q-learning) will be applied to the extended data set in order to find an optimal strategy for placing the bets over regular NHL season in order to maximize the net profit.

From HPC to mobile development, the prevalence of accelerators and other performance-driven architectures is a fact you can't argue with anymore. What if you want to tap into those source of performances but you don't really want to sacrifice the elegance of your C++ 20 ? What if you may want to explore some of those architectures now but change your mind later without dropping all the code you already wrote ?That's where SYCL comes down. SYCL is an open standard aiming at providing a cross-platform programming model, tools and compilers to target accelerators at large. Made to be interoperable with C++ , it is simplify the design, debugging and deployment of applications over a large selection of accelerators: multi-cores, GPGPU or even FPGA.In this talk, we will give a short tour of SYCL saillant point, how to get started with it, how it can smoothly be integrated in your C++ code without major changes in your programming habit and we will conclude with some result of such an integration in a High ENergy Physic applications.

Jeff & Amin describe how Google is using ML models in a variety of ways to provide services to their customers. Jeff’s portion of the talk focuses on software and the models that they are using, while Amin’s section describes the customized, highly scalable hardware that Google is building in order to allow these models to function at scale.

In the latest SBC Podcast I was joined by Jason Trost – the founder and CEO of both the betting exchange, Smarkets and the bookmaker, SBK.

With so many incredibly important topics up for debate in the betting world at the moment, Jason was the perfect guest to explore concepts like affordability checks and their considerable impact on those of us that enjoy betting.

Jason also talked openly and candidly about the reality of running both Smarkets and SBK, trying to do things differently to their (admittedly bigger) competitors, exchange liquidity, commission and much more.

He also outlined how SBK has been setup to offer better odds than many of their rivals and just how it does that, plus his goal to make it appeal to punters looking for a more traditional bookmaking experience.

Sure, graphs can be horrible and convey no information or even actively mislead. Tables can be hard to read, but I guess that, without actually faking the numbers, it would be hard to make a table that’s as misleading as some very bad graphs.

To avoid becoming next Zillow, when hiring #datascientists for time-series related jobs data science teams should exercise caution when hiring self-declared “experts” in time series forecasting.

I continue to be bothered by the term “p-hacking.” Sometimes it applies very clearly (as in the work of Brian Wansink, although it’s a mystery why he felt the need to p-hack given that it seems that his data could never have existed as reported), but other times there is no “hacking” going on. So I prefer the term forking paths.

Probability estimation models play an important role in various fields, such as weather forecasting, recommendation systems, and sports analysis. Among several models estimating probabilities, it is difficult to evaluate which model gives reliable probabilities since the ground-truth probabilities are not available. The win probability estimation model for esports, which calculates the win probability under a certain game state, is also one of the fields being actively studied in probability estimation. However, most of the previous works evaluated their models using accuracy, a metric that only can measure the performance of discrimination. In this work, we firstly investigate the Brier score and the Expected Calibration Error (ECE) as a replacement of accuracy used as a performance evaluation metric for win probability estimation models in esports field. Based on the analysis, we propose a novel metric called Balance score which is a simple yet effective metric in terms of six good properties that probability estimation metric should have. Under the general condition, we also found that the Balance score can be an effective approximation of the true expected calibration error which has been imperfectly approximated by ECE using the binning technique. Extensive evaluations using simulation studies and real game snapshot data demonstrate the promising potential to adopt the proposed metric not only for the win probability estimation model for esports but also for evaluating general probability estimation models.