Computer Science: KESS II Funded MSc by Research Studentship at Swansea University: Adopting data driven modelling and prediction approaches to support strategies for successful game outcomes in Rugby
- Full cost of UK/EU tuition fees, plus a stipend
- 31 August 2017
Swansea University is a UK top 30 institution for research excellence (Research Excellence Framework 2014), and has been named Welsh University of the Year 2017 by The Times and Sunday Times Good University Guide.
*This scholarships is part funded by the Welsh Government’s European Social Fund (ESF) convergence programme for West Wales and the Valleys.*
MSc by Research for UK or EU applicants in the field of Computer Science / Visual and Interactive Computing
This work in with Ospreys Rugby will consider the design of Key performance indicators (features) and a Bayesian-based machine learning models to predict the outcomes of a rugby match. The approach will focus on the ability for the above data-driven features and models to inform coaches and players on areas that are mostly likely to influence game outcomes therefore help shape training, selection and strategies in Rugby Matches. This research will empower Ospreys Rugby and the wider community (through it is community initiatives) to further the adoption of data driven techniques to enhance performance and gain competitive margins.
The work has four main research objectives:
1) The design and quantifiable validation of in-game context-sensitive data driven descriptive statistics.
Data driven analysis of the key performance indicators and their inferred measurable influence on game outcomes are an important tool for professional Rugby teams. However, lots of the current state-of-the-art post-match key performance indicators of players are aggregated whole game statistics. This work will seek to design in-game context-sensitive key performance indicators and their quantifiable impact on game outcomes. For example, passing a certain amount of time during a game may be a desirable tactic, however kicking at the right moment in the given context of the game may be better. Of course, this is understood by coaches and players alike but not well embedded in to the post-match descriptive statistics which are used to review performance. Therefore, this work will seek to design new indicators (features) that embed the context of the game.
2) To build a predicative model with some transparency of feature impact on outcomes.
Predictive modelling and Machine Learning is a growth area in data science and has been applied in several ways to sport. This work will build Bayesian-based statistical models to use in understanding in-game performance and prediction of game outcomes. However, it is often the case that such complex models can gain in performance (game outcome prediction) but lack transparency or interpretability. This work will seek to adopt RNN and RBM Deep learning techniques that will model the set of interconnected event sequences and their influence on the outcomes metrics (described in next section). Furthermore, when incorporating additional, multimodal data, such as GPS or biometric data, an ensemble modelling approach will be adopted in order to maintain actionable inerrability to the models whilst maintain performance capabilities of state-of-the-art deep learning approaches.
3) This work will adopt and investigate the relative usefulness of continuous outcome metrics to train supervised probabilistic predication of scores.
The work will also investigate the use of proxies to train such models such as field position (or gains in), possession (time in possession, turn overs of possession) and finally momentum. Momentum in sport games is contentious metric that many people talk about as a proxy indictor of current as predicated success of a team during a sports match but there is some scepticism that this is real or just inference of a pattern that is not really in the data.
4) The work will assess the usefulness of the features and predictive models trained using various outcome metrics in empowering the team to investigate and make data-driven changes to training, selection and in-game strategies.
Scholarships are collaborative awards with external partners including SME’s and micro companies, as well as public and third sector organisations. The scholarship provides 1 year funding with a 3 month period to complete the thesis. The achievement of a postgraduate skills development award, PSDA, is compulsory for each KESS II scholar and is based on a 30 credit award.
Candidates should have a 2.1 or above in their undergraduate degree Computer Science or a related subject. They should also be eligible for UK/EU Fees.
The studentship covers the full cost of UK/EU tuition fees, plus a stipend. The bursary will be limited to a maximum of £11,472 p.a. dependent upon the applicant’s financial circumstances.
There will also be additional funds available for research expenses.
Please visit our website for more information.
Sign Up for Scholarship Updates
Get an email every week that 10.000's of students use to get the latest scholarships.