Visualisation code for my winning entry to the Liverpool Analytics Challenge by Friends of Tracking. Optimised for x6 speedup.
LINK TO ANALYSIS, on Medium
VIDEO OF PRESENTATION, on FoT Youtube channel
Pitch Control measures the probability that a team will retain possession of the ball if they pass it to another location on the field. It can be used to evaluate passing options for a player, and quantify the probability of success.
This repo is for recreating the Pitch Control videos shown above, with a x6 speedup1 from using Python's joblib
.
To get started, clone this repo, including submodules2 using:
git clone --recurse-submodules https://github.com/suryako/FoT-Liverpool-Analytics-Surya.git
Install requirements and start a jupyter session with:
pip install -r requirements.txt
jupyter lab
Open main_notebook.ipynb
and run the cells in the notebook sequentially.
The Pitch Control videos will be generated at goals_pc/_____.mp4
The notebook will allow you to interact with the data to get a better feel for it.
Many thanks to David Sumpter and Friends of Tracking for organising this, Ricardo Tavares for providing the dataset, and Laurie Shaw for the visualisation library.
1
x6 speedup on quad-core MacBook Pro with 8 vCPUs, increase n_jobs
accordingly if you have even more CPU cores.
2
The dataset by Ricardo, and Ciaran's data format converter are included as submodules, locked at specific commit tags for reproducibility.
Laurie's code (and Pitch Control model) is directly included in this repo, to keep track of some of my changes that are specific to the Liverpool goal data.