AFL match review using GPS data
Project Background
Understanding how workrate and high speed efforts change throughout a match is essential for high-performance staff when assessing team endurance and individual player output. This project focuses on two dashboards designed to analyse work rate, high-speed running, and sprint running across each quarter and compare match performance to season averages.
These metrics are measured in metres per minute (m/min) to account for variations in match and quarter lengths, as well as differences in player “time on ground”. Using absolute metres would not provide an accurate reflection of intensity, as players with lower game time would naturally record lower totals.
The first dashboard provides a quarter-by-quarter breakdown for the entire team, highlighting performance drop-offs over a match and how match metrics compare to season averages. The second dashboard focuses on individual player performance, displaying their match metrics against their personal best. High-performance staff can use these metrics to assess performance and determine whether athletes are maintaining their intensities throughout the season. Coaches can evaluate individual player output, and in cases where selection changes are needed due to omission or injury, they may prioritise a player with a similar physical output to ensure a like for like replacement.
Stakeholders
High-Performance Manager
Athletes
Strength & Conditioning coaches
Physiotherapists
Selection Committee (Head Coach, Assistant Coaches, Head of Football)
Data Structure
This is a real-world dataset taken from Australian football athletes across an entire season. The datasets were randomised to protect privacy and intellectual property. 4 tables were used, consisting of the following structure:
quarter_by_quarter: 836 rows x 6 columns
fixture: 9 rows x 3 columns
profiles: 29 rows x 2 columns
afl_teams: 19 rows x 2 columns
Data Cleaning, Transformation & Manipulation
This project focused on extracting, transforming, and analysing GPS match performance data to assess team and player work rate trends. The SQL workflow incorporated common table expressions, joins, aggregations, and ranking functions to process and structure the data for visualisation.
Team averages for work rate, high-speed running, and sprint running per match were calculated using aggregate functions. Percentage change and binary calculations were applied to highlight positive and negative shifts relative to the season average.
Match performance data, broken down by quarters, was merged with fixture details to align match rounds with opponents and dates.
Pivoting techniques were then used to restructure the data into a format suitable for visualisation tools.
Subqueries identified each player’s personal best performance before their current match, and percentage change calculations measured improvements in work rate, high-speed running, and sprint running. Rankings were applied to highlight the players who showed the most improvement in each match across these three metrics.
Data Visualization
Power BI was used to visualise the change of intensity each quarter, incorporating filters to select specific rounds with opposition logos linked. Binary performance indicators were used to display up or down arrows when comparing match performance to the season average, requiring custom columns to determine colour selection. The player dashboard leveraged ranking data to highlight the most improved players, with profile images dynamically retrieved using the image URL.
For this project, the scenario question is used:
"What were the running outputs like in the Round 8 match?"
Overview of findings
The squad’s overall workrate, high speed running, and sprint running were all below the season average in Round 8, with a particularly slow start in the first quarter. While a few players, such as Janik Sinner and Elon Musk, maintained performances close to their personal bests, many struggled, with sprint running heavily impacted by the wet conditions. Intensities improved as the match progressed but dropped again in the final quarter, where workrate fell to 103 m/min, high-speed running to 24.1 m/min, and sprint running to 3.4 m/min.
Squad Level Findings
In Round 8, the squad’s workrate was 8% lower than the season average, with high-speed running down by 12% and sprint running down by 10%.
The team had a slow start in the first quarter, recording 104m/min for workrate. Typically, the first quarter sees the highest intensities across all three metrics, yet high-speed running was 26.7 m/min (the second lowest quarter of the match), while sprint running was just 3.3 m/min (the lowest quarter of the match).
As the match progressed, intensities across all three metrics improved before declining again in the final quarter, where workrate dropped to 103 m/min, high-speed running to 24.1 m/min, and sprint running to 3.4 m/min.
Player Level Findings
Janik Sinner and Elon Musk were the only two athletes to achieve within 10% of their personal best for work rate, recording 116m/min and 115m/min respectively. Alexander Zverev showed the greatest high-speed running in comparison to personal bests, finishing just 7% short of his at 33m/min.
Donald Trump recorded the highest workrate (123m/min) and high-speed running (41m/min) for the match, while also setting a 4% personal best in sprint running, reaching 7.6 m/min.
Several players recorded less than 50% of their typical sprint running output, largely due to continuous rain, which impacted ball movement and led to frequent stoppages, reducing overall match intensity.
