Towards predicting optimal team lineups
Master’s thesis/Research internship, M2, 6 months, 2023
Laboratory/company
Laboratory: GREYC CNRS UMR 6072
Team CODAG – Contraintes, Ontologies, Data mining, Annotations, Graphes
Université de Caen Normandie
14000 Caen, France
Company: Skriners
38 rue de Metz
92000 Nanterre
Remuneration
Standard gratification for a research internship according to French
legislation: ~570 euros/month
Context
Using computational methods to analyze sports data gives practitioners
(coaches, agents, athletes themselves) powerful tools to make more
objective decisions when it comes to a variety of questions that arise in
professional sports.
The company Skriners already offers a tool for supporting decision makers
for player acquisition or replacement, based on sophisticated statistics
derived from video recordings of matches. Skriners is a SaaS software for
sports professionals to compare, recommend and manage players based on
statistical criteria. Thanks to its comprehensive database, Skriners can
also help find promising talent. The software also offers a workforce
management feature. This decision support is limited to individual players,
so far, not taking teammates or eventual information about opponents into
account.
In the long term, the tool is to be enriched to automatically suggest team
lineups, based on available players, intended match strategy, information
about the opposing team etc. This will require taking synergies between
players into account, as well as the performance of particular players in
given defensive or offensive systems.
The work to be performed in this internship will lay the groundwork for
this future research, by exploring whether and how existing work on team
chemistry [1], the context of players’ performance [2], and the automatic
identification of tactical formations [3] can be applied to the data
currently available to Skriners. Based on this evaluation, the intern will
either start implementing and applying those techniques to the data to
derive additional statistics, or identify in which way data and/or methods
need to be adapted.
Objectives
- Evaluate the applicability of existing methods to the data available to
Skriners - Evaluate the needs for and possible sources for additional data
Activities
- Familiarize oneself with the data at Skriners’ disposal
- Familiarize oneself with existing work in the literature
- Identify whether there are data that would be needed but are currently
missing - Implement and apply existing methods to the data, generating additional
statistics - Identify additional data sources
Profil
Student in computer science or sports/movement science. Programming, as
well as machine learning/data mining or statistics knowledge necessary.
Candidates are encouraged to apply as soon as possible.
To apply
Send the following documents (as .pdf) to the contact addresses listed
below:
- Motivation letter
- Résumé
- Grades for M2 (as far as available) and M1
- If possible, contact data for one or more persons (teachers, prior
internship supervisors) who can be contacted for references
Contact
albrecht.zimmermann@unicaen.fr
gill.affoum@skriners.fr
References
[1] Bransen, Lotte, and Jan Van Haaren. “Player chemistry: Striving for a
perfectly balanced soccer team.” arXiv preprint arXiv:2003.01712 (2020).
[2] Bransen, Lotte, Pieter Robberechts, Jesse Davis, Tom Decroos, Jan Van
Haaren, Angel Ric, Sam Robertson, and David Sumpter. “How does context
affect player performance in football?.” (2020).
[3] Bialkowski, A., Lucey, P., Carr, P., Yue, Y., Sridharan, S. and
Matthews, I., 2014, December. Large-scale analysis of soccer matches using
spatiotemporal tracking data. In 2014 IEEE international conference on data
mining (pp. 725-730). IEEE.