Can Data Influence Recruitment At Championship Football Level?
- 31st January 2019
- David Hall
At Peak Indicators, we’ve been working with Paul Clough for many years. This recently led to Paul taking up the role of Data Science Lead with us, alongside the role he holds with Sheffield University. It was also partly this relationship that led to a collaboration between the following 3 parties:
- Ross Burbeary, a football expert from Rotherham FC, with a stack of data to hand.
- A group of Sheffield University students with an interest in football and a dissertation to write.
- Our team at Peak Indicators, since we had a real desire to prove the architecture of Azure and Power BI (more on that later).
Initially, Ross met Paul and some of his Sheffield University students during one of their regular meetups; where like-minded individuals look to find ways to get the most out of data and adopt advanced analytics; to improve decisions around recruitment.
What Did Rotherham FC Need?
Rotherham FC receive weekly data feeds via technology provided by Instat. Ross, however, felt that there would be an easier way to get the full insight they needed from this data. He questioned whether a criteria-based approach to utilising data and KPIs could be created for the purpose of scouting players.
He ultimately wanted to discover whether there was a way of using advanced analytics to scout a player that’s demonstrated the capacity to fulfil a role in the Championship division, even if this player is playing leagues below or abroad.
Furthermore, Ross wanted answers to the question: If a player gets injured, could we replace him, via KPI data in that position related to a new signing? Essentially, can data influence recruitment at Championship football level?
What Was the Student Involvement?
Out of this slightly vague hypothesis, one of Paul’s MSc Data Science students, David Jerome, adopted the following dissertation title: Investigation of Instat Football Dataset - Player Clustering & Classification.
The full dissertation can be accessed upon request, but it was clear that it was indeed possible to classify and cluster players. There were, however, challenges to overcome for this to be achieved in a timely and repeatable manner; via an architecture that could evolve over time.
It was at this point that Paul looped in our team at Peak Indicators to turn their engineering skills to David’s thesis.
Peak Indicators Involvement
There were a number of challenges for us to overcome to assist:
- The data needed cleansing, which had to be done quickly and automatically so that weekly match data could be made available as soon as the raw data was presented.
- We needed to explore additional algorithms against the improved data, on a more powerful platform; leveraging our in-house data science team and our Tallinn platform to build upon David’s findings.
- We needed the results to be presented in an interactive, visually-compelling way, so that end users could use the data to augment their decisions.
After establishing these aims, the initial meetings we had coincided with Peak Indicators becoming a Certified Data Analytics Partner for Microsoft. What better way to train more of our team on the Azure and Power BI platform than moving the theory into production practice?
Carrying Out the Work
We carried out the work required from our Chesterfield-based hub, where we have access to 40 analytics professionals. To complete the project, we used the following staff:
- Azure Architect: to build a stable, robust and future-proof architecture.
- Power BI Expert: to build an easy-to-use, set of role-based dashboards.
- Data Scientist: to assist with access to our Tallinn platform.
Using the power of the Azure platform, the Peak Indicators team automatically ingested and cleansed the data using Data Flow, then wrote the results to an Azure SQL database. We then used our automated machine learning platform, Tallinn, to automatically find the best algorithms.
We were then able to present these augmented results via an interactive set of Power BI dashboards. These have been really positively received.
The Next Steps
Like any analytics programme, the solution will continue to evolve. The Azure platform and robust design means we are able to quickly tailor dashboards to more roles or key areas of insight.
Moving forward, we now also expect to add more data sources; which will only enrich the results. We will then continue to work with more teams, who will provide more feedback.
Contact Peak Indicators
Interested in finding out how to leverage the power of our Tallinn machine learning and methodology to the benefit of your own business? Contact us today.