-
An example dashboard (Millwall)
Inspiration
After discussing the available problems to tackle, we were particularly intrigued by the challenge and limitations of working with a provided, previously-unseen dataset. This challenge seemed to provide us with the best opportunity to develop our technical skills in web development and data processing whilst providing us with the opportunity to realise a tangible output from our efforts, in the form of useful insights. It was also interesting to learn about the uses of technology within a sport that, from the outside, didn't seem to be particularly able to benefit from modern technology.
What it does
Our project takes a dataset with statistics about matches from the past 4 seasons of the EFL Championship it then cleaned the data and ran some statistical analysis on the data. We used the analysis to make suggestions about game tactics and training as well as some statistical summaries. Finally, we put it all together into a graphical webface, where it would be accsesible for anyone within the club for instance a coach to interact with as well as making sure the information modern and user friendly. This means the information can be looked at quickly whilst being able to gain insight from it even as a novice to data science.
How we built it
We build the platform around a Django project, which provided a comprehensive and adaptable way of constructing a basic web server, which was especially helpful as we continued to develop the project, changing our requirements, desires, and subsequently changing our django project. To support the back-end processing, we wrote a script that breaks down the large data set provided into smaller csv files, using pandas, which contain all the matches played against a single competitor. This allowed us to do in-depth analysis focusing around Coventry City FC's performance against individual competitors, rather than just focusing on their over-all performance throughout the season. This analysis was supported by python scripts that read and manipulated the data from these CSV files into formats that could provide useful insights to the coaches, decision makers, and team-members at Coventry City FC.
To improve the accessibility of these insights, we used HTML (specifically jinja templates), CSS, and JavaScript to convey this information in an easy to digest format that is adaptable to different insights provided by our analysis of individual competitors.
The development of both the front and back end underwent several substantial iterations, from a basic re-printing of statistics interpreted from the provided data set to technical analysis of deconstructed data-sets that provided focused insights conveyed through clear graphs, charts and tables.
Challenges we ran into
The inital challenge we ran into was in breaking down and processing the large dataset. How we overcame this was by cleaning the data and splitting the dataset into sub sets by opposition team. This ended up being beneficial as we were able to make our insights specific to each team and adapt suggestions based on that. Another challenge was in formatting the front end so that it was dynamic meaning the original dataset could be added to and our webpage would update accordingly. This meant our graphs and text insights needed to be capable of adapting automatically. We also tackled this project with tech stacks that we were unfamiliar with, which inhibited our inital progress as we learned it, but once we got started we managed to pick up a lot quickly. Finally, we adapted our solution as we learnt more about the data and the interesting statistics we learnt . This was a challenge as our plan changed during the intial stages of the project.
Accomplishments that we're proud of
Finding new ways to manipulate data with statistics in python and applying it within the context of a Football Club. Using chi-squared distribution to perform hypothesis testing on whether Coventry F.C. were following through with expected goal predictions as well as if they overperformed in defending against goals were the expected was higher than the actual conceded goals. We are also proud of the networking with our peers and finding mentors in the voluenteer's that were with us throughout the whole event.
What we learned
We learnt lots about team communication and ensuring that you are checking in with team mates both to montitor their progress as well as making sure no one felt overwhelmed with the tasks. We gained a lot of resilience from code that wouldn't always work as well as the mental barrier of pushing through for many hours in a row working on the same project.
What's next for Theodore's Tech Titans
In the future it would be helpful to have access to a more comprehensive data set, so we can identify more long-term trends, and provide more accurate insights. There is also room to further improve the functionality of the front end by providing sorting and filtering options, allowing comparisons between multiple competitors on one data set, and introducing predictions for future games using previous trends and probabilities.
Log in or sign up for Devpost to join the conversation.