Throughout the autumn of this year, OptaPro Analytics Forum judges faced another difficult challenge in selecting the presentations and posters for the 2019 event, which takes place on 6th February in central London.
Entering its sixth year, the OptaPro Forum remains a key date in the football analytics calendar, attracting over 60 clubs from around the world.
The full line-up is now confirmed and can be read below (presentations are listed in no specific order).
Ben Torvaney – Player2vec: A sequence-first approach to player and team styles
This project aims to enhance our understanding of playing style. By analysing where in a possession sequence a player tends to contribute, we can build a quantitative representation of their playing style. Comparing and aggregating these representations at scale has applications in high-level scouting and tactical analysis.
Ben is a data scientist based in London. After two years working in sports betting and analytics, Ben now works at Wefarm, the world’s largest farmer-to-farmer digital network.
Ben also presented at the 2018 OptaPro Analytics Forum, and a write-up of his presentation can be read here.
Carlos Rodriquez – Heading the wrong way? Estimating body orientation in professional football
Coaches refer to the 2D orientation of the player’s upper-torso as a key feature for obtaining a positional advantage in the game. Orientation data is expected to be an insightful feature for fine-tuning spatio-temporal models for tactical analysis such as pitch control, pass probability or defensive pressure. Carlos’ presentation discusses a complete computer vision pipeline for predicting the orientation of soccer players from panoramic videos.
Carlos works in the sports science department at F.C. Barcelona, where he provides key support and insight to aid coaches in their everyday analytical tasks through computer vision and machine learning. His current focus surrounds merging event and positional data to develop algorithms informing tactical concepts. Carlos is a computer scientist and holds a master’s in Innovation and Research of Informatics.
Andrew Rowlinson – Perfecting your set piece technique: finding an edge to glory
Andrew’s presentation applies official LaLiga tracking data to identify an opponent’s defensive vulnerabilities from set piece situations and outline specific routines that exploit these weaknesses.
Andrew Rowlinson is a machine learning specialist with a decade of experience in working with data and predictive modeling. He currently works at Bluugo, using data science to digitalise and improve logistics supply chains.
Joe Mulberry – Gazing into latent space to find an edge with possession sequences
The dynamic nature of football could be better described using sequences of actions rather than singular actions. Analysing sequences using a combination of an auto-encoder and T-SNE, will allow their categorisation using their location within latent space. The categorisation will allow analysis of team dynamics as well as providing further context to player identification processes.
Joe has worked in player recruitment for the more than 15 years across Africa with the Right to Dream Academy and in Europe with FC Nordsjaælland. Joe also presented at the 2017 OptaPro Forum.
Mathieu Rosenbaum and Othmane Mounjid – Optimal coaching: How the tools of artiﬁcial intelligence and data science can help
This presentation showcases a new methodology to inform and enhance coaching decisions, applying artificial intelligence concepts more commonly associated with the field of quantitative finance. Through simulation and the theory of Markovian stochastic control, the presentation discusses optimal pre-match and in-game strategies to support head coaches.
Mathieu Rosenbaum is a professor in the applied mathematics department at École Polytechnique in Paris, where he holds the chair ‘Analytics and Models for Finance’. He has published more than 60 articles in data science, applied probabilities and financial mathematics.
Othmane Mounjid graduated from Ecole des Mines in applied mathematics and engineering in 2015 and from Sorbonne University in probabilities and quantitative finance in 2016. He is now a PhD student at École Polytechnique in Paris under the supervision of Mathieu Rosenbaum, working on financial markets modelling and optimal trading.
Mladen Sormaz and Dan Nichol – Quantifying the impact of off-the-ball movement in football
Through applying official LaLiga tracking data, this presentation aims to characterise off-the-ball runs by first examining the `damage’ they induce in opposing team structure. By correlating this damage statistic with well-established metrics, this work aims to derive a model of run quality that can be used in scouting, opposition analysis or player development.
Mladen holds a PhD in Cognitive Neuroscience and currently works as a full-time data scientist at Football Whispers. He also works as a part-time first team data analyst for Huddersfield Town where he provides tactical analysis based on event and tracking data with the video and tactical analysis teams.
Dan holds a computer science DPhil and currently works as a postdoc at the Institute of Cancer Research. His research aims to build predictive models of cancer progression to aid in the design more effective treatment. In football analytics, Dan uses the same prediction techniques to identify key game moments and derive partially automatic performance analysis tools.
Jakub Michalczyk – A glance at building out from the back
Jakub’s study focuses on clustering passes made deep on the pitch by defenders or goalkeepers that push the ball higher, aiming to detect any trained schemes or tactical assumptions that determine the initial stages of each attack. Such work could inform a tactical analysis and help to plan how to stop opponent’s attacks in the initial stage or, just by a team positioning, impel an opponent to play the ball in an untrained or inefficient way.
Kuba is a mathematics graduate, currently working as a data scientist in a media agency. Earlier this year, Kuba gained a place on Dan Altman’s North Yard Analytics fellowship scheme.
Michal Jaron, Extended ghosting team model – a tool for planning and testing actions
Recent advances in reinforcement learning enable us to simulate multi-agents’ behaviour on an unprecedented scale. This poster presentation applies official LaLiga tracking data, showcasing real-time analysis and in-depth application of the ghosting approach.
Michal is currently finishing Computer Science studies at University of Warsaw and started his football journey at Legia Warsaw, where he was responsible for underpinning foundations of data-driven automatic analytics and opposition analysis.
Tom Brown – Playing style interactions and optimising player synergies
Tom’s research extends on a project from last year’s Forum by Mark Carey and Mladen Sormaz to first identify playing styles within positions and then classifying each player into a playing style. Following this Tom will then examine how a players’ performance is dependent on the interrelationships and synergies between their and their teammates playing styles.
In his day job, Tom is a data scientist working in the defence and cyber security industry.
OptaPro would like to thank all who submitted a proposal as well as the judges, and congratulate the 11 people who will be presenting or showcasing a poster at the 2019 OptaPro Analytics Forum.