Following the announcement last month of the first projects chosen for the 2023 Opta Forum, a further four proposals, from the research competition’s Opta Vision track, have been chosen for presentation to an invited audience of industry practitioners in London later this month.
Shortlisted participants had an opportunity to interrogate Stats Perform’s most sophisticated performance metrics, which include unique metrics trained on player tracking data, from 100 matches played in the 2021-22 Ligue 1 Uber Eats season. A full list of Opta Vision metrics can be found here.
Alongside the Opta Vision metrics, the candidates also had access to official tracking data from the same matches in the dataset.
Each candidate submitted a notebook, to showcase their analysis and method, together with a short video and PowerPoint presentation, which were evaluated by Opta Forum judges.
The projects selected in the Opta Vision track of the 2023 Opta Forum, listed in no specific order, are as follows:
Hugo Rios-Neto – Characterising Off-the-Ball Spatial Occupation
Hugo’s project proposes the creation of a new algorithm, ‘SoccerRep’, which is trained on Opta Vision data to learn complex, low-dimensional representations, from multiple layers of fine-grained spatio-temporal data in football.
SoccerRep can be trained on any pitch surface, with its choice depending on what the objective of the analysis is. For this project, he trained SoccerRep on Individual Pitch Control surfaces to allow for the characterisation of the off-ball playing styles of different players. When used in the context of match analysis, it can help analysts better understand the spatial behaviour of their own team and of their opponents. In scouting, this enables teams to search for similar players based on how they occupy spaces in both attacking and defending scenarios.
Hugo is currently completing an MSc in Computer Science at the Federal University of Minas Gerais in Brazil, whilst also working in football as a data scientist with Serie A side Atletico Mineiro. He was supported in his research for this project by Professor Jesse Davis and Maaike Van Roy, who are both based at KU Leuven, as well as his supervisor at UFMG, Wagner Meira Jr.
Jack Pamukci – Tackling the Maldini Doctrine: Quantifying Defensive Positioning
Jack’s presentation aims to propose a method for measuring how a team out-of-possession can limit their opponent’s progressive play in build-up phases and counterattacks, through assessing the positioning of their players on the pitch.
His approach utilises a combination of Opta Vision data points, pitch control, and predictive modelling to create two new metrics: Generalised Defensive Presence (GDP) and Dangerous Recovery Value (DRV). The application of these two metrics will provide coaches and recruitment staff with a more objective understanding of how each individual player defends for a team, as well as quantify how proactive a player is in looking to prevent opposition attacks from developing.
Jack is studying Computer Science at the University of North Carolina at Charlotte, where he has also spent six months as an Undergraduate Teaching Assistant within UNC’s College of Computing and Informatics. He also has experience as a Full Stack Developer for Toyota Racing Development, as part of another collaboration with the University.
Daryl Dao – Using Data to Evaluate the Efficiency of Channel Attacks and Defence
Modern attacking positional play generally comprises five vertical lanes across the pitch: left flank, left half-space, centre, right half-space and right flank. The intention of attacking all five lanes on the pitch is to stretch, and hopefully open up space, between defenders. While central areas are usually more congested defensively, full backs (or wing backs) can often be drawn out to the flank, potentially opening up a large channel between them and the nearest central defender.
Daryl’s poster aims to utilise Opta Vision data to quantify and evaluate channel attacks, from both attacking and defensive perspectives, with the objective of attaining a better understanding of how teams can increase their efficiency. In attack, he will look to establish how channel attacks are created and how much danger can be created from them, and in defence, how formations, structures and non-defender support can defend against these attacks.
Originally from Vietnam, Daryl is now based in Queensland, Australia, where he is studying Computer and Data Science at the State’s University of Technology. In parallel to his studies, he has also worked as a data analyst for Queensland NPL side Olympic FC.
Severiano Jimenez, Will Bantz and Patrick Bardsley – Threat to the Future: Understanding the Importance of Expected Threat
Severiano, Will and Patrick will be displaying a poster which presents the key findings of a project which evaluates the relationship between Opta Vision metrics, with an emphasis on Expected Threat, the predictive value of the metrics on the final result of a game and the likelihood of a player having a significant impact on a game, based on their season performance in these metrics.
The research project is influenced by their own experiences of how coaches, working for their collegiate team, were able to successfully implement adjustable KPIs into their coaching sessions and influence on-field performance.
Severiano, Will and Patrick are all members of the analytics team working for the Charlotte 49ers, UNC Charlotte Men’s Soccer team, who compete in The American conference of NCAA Division I. Severiano is studying for a BS in finance and business analytics, Will is studying for a BS in finance, and Patrick will be completing a computer science degree (BS) in May.
As was the case with participants in track one of the competition, Stats Perform would like to thank everyone who submitted a proposal in the Opta Vision track and congratulate the four groups who will be presenting or exhibiting at the 2023 Opta Forum.
In addition to the presentations in the Pro research track, the Opta Forum will also feature a range of panel discussions and guest talks from leading industry figures, working across the Pro sector as well as in the media space, providing insights into how applications of data can inform performance, decision-making and wider fan engagement.