Following the deadline for submissions, the OptaPro Analytics Forum judges spent November reviewing dozens of presentation proposals for the 2020 event, which will be taking place in central London on 5th February.
The judges scored each proposal, which were anonymised, based on three key criteria: innovation, relevance and application. By the end of the process, six projects had been chosen for presentation, with a further two submitters invited to exhibit their proposals as posters.
Now entering its seventh year, the Forum remains a key date in the football analytics calendar, showcasing innovative approaches to key questions relating to performance analysis and recruitment to industry professionals working at over 80 clubs and federations worldwide.
The full line-up for the 2020 Forum, listed in no specific order, is as follows:
Karun Singh – Learning to watch football: Self-supervised representations for tracking data
This project investigates a novel approach to identifying recurring match scenarios using tracking data, with the aim of enhancing a performance analyst’s existing working processes. A frame of tracking data only describes the location of each player and the ball, but this can be used to derive other meaningful information about the frame (e.g. Voronoi diagrams) for free, without any need for human data labelling. By leveraging such ‘self-supervised’ labels, we can learn representations that help us quickly identify similar scenarios across matches without having to scour through hours of video.
Based in San Francisco, Karun is a software engineer working in the Applied AI space. He graduated from Cornell University’s College of Engineering in 2018, where he was recognized as a Merrill Presidential Scholar. He majored in Computer Science with a focus on computer vision and machine learning.
David Quartey – How do attacking and defending strategies affect goalscoring opportunities from throw-ins?
During the 2018/19 Premier League season, over 22 throw-ins occurred on average each game. However throw-ins remain a relatively undervalued set piece opportunity for generating goalscoring opportunities.
Using Opta data, David’s presentation will outline the findings of a logistic regression model to highlight the different types of attacking throw-ins used by teams in the final third, and the likelihood of them resulting in goals, shots or big chances against different types of defending. It will also draw conclusions on whether a mixed throw-in strategy works better than a consistent throw-in strategy in terms of generating goalscoring opportunities.
Originally from Ghana, David is an economics graduate who recently gained a place on Dan Altman’s North Yard Analytics fellowship scheme in New York.
David Perdomo Meza, Daniel Girela and Mark Thompson – Tactical insight through team personas
This project was chosen in the Swansea City-led submission category, which focused on identifying playing styles which are effective against certain formations or playing styles in the EFL Championship.
Building on a ‘team personas’ concept originally showcased by David on a poster at the 2018 Forum, this project features an applied analysis of the relationship between formations utilised, stylistic team personas and the final match outcome. The presentation will also focus on personas at player level and a player’s aggregated match contributions against different formations, which can inform team selection.
David and Daniel both work as full-time data scientists for Twenty3, alongside Mark who is also employed there as a data analyst.
Santhosh Narayanan – Modelling event sequences in football using multivariate point processes
Santhosh’s presentation applies Opta data to model on-the-ball event sequences with the objective of building a real-time match simulator, to generate performance predictions at player and team level which can inform in-game decision making.
Building on the Hawkes process mathematical model, Santhosh’s model incorporates multivariate point processes to establish the relationship between one event in a possession sequence and the likely outcome of the next, which in turn can help predict the occurrence of events leading to goals.
Santhosh possesses an MSc in Data Science at the Barcelona Graduate School of Economics and is currently completing a PhD at the University of Warwick. He is also a visiting student at the Alan Turing Insititute in London.
Dan Barnett – I’m in a wide open space: creating opportunities at set pieces
From studying tracking data across an entire season, Dan’s presentation identifies player movement strategies prior to a corner kick or free kick set piece, as well as movement after the ball has been delivered, to establish potential repeatable strategies that can generate high-quality goalscoring opportunities.
Dan is the Director of Analytics at Analysis Marketing, providing analytical insights relating to customer behaviour for clients across different sectors, including Telecoms, Media and Financial Services.
Vignesh Jayanth – Identifying and evaluating strategies to break down a low-block defence
In the second of the Forum’s club-led submission categories, Vignesh Jayanth’s presentation focuses on strategies for breaking down a low-block, with a specific emphasis on the tactical style of Danish Superliga side FC Nordsjælland.
Through applying tracking and event data, Vignesh’s project summarises sequences in possession chains against a low-block, tactically relevant to Nordsjælland. Based on these insights, he will outline strategies for breaking the low-block through a combination of horizontal and vertical movement across different pitch zones in the attacking half.
Vignesh is currently studying for an MSc in Data Science besides working as a part-time in-play football trader. He has previously worked as a first-team performance analyst at a League One football club and has obtained PG Diploma in Sport Coaching.
Pieter Robberechts – Contextualized performance projections for soccer players
By applying Opta data through a deep neural network, this poster presents a recruitment model for assessing a potential signing’s ‘fit’ by forecasting his future performance at the club. Therefore, the model learns how a player’s playing style and strengths align with the team’s current tactics and playing personnel.
Pieter a PhD student in the DTAI Research group at KU Leuven, where his research interests include the application of artificial intelligence, data science and machine learning in sport.
Javier M. Buldú – Using network science to quantify the identifiability of football teams
The application of Network Science to a variety of fields, from social networks to brain dynamics, has demonstrated the approach’s versatility in analysing complex problems from a new perspective. Using this this approach, Javier’s poster looks to quantify to what extent the playing style of any team is maintained over the course of a season, based on the persistence of passing patterns.
Possessing a PhD in Applied Physics, Javier is the coordinator of the Complex Systems Group at the King Juan Carlos University in Madrid, as well as being the Principal Investigator of the Laboratory of Biological Networks at the Center for Biomedical Technology.
OptaPro would like to thank everyone who submitted a proposal as well as the panel of judges, and congratulate the 10 people who will be presenting or showcasing a poster at the 2020 OptaPro Analytics Forum.