Inspiration
We were initially inspired by doing a project with the Women's National Team. Our first idea was to utilize Men's FIFA rankings to create characteristics of each position. Once we had the characteristics of what makes a good defender, forward or mid-fielder, we were going to apply that to the Women's National Team data to create rankings similar to that of the FIFA rankings for men. After creating those rankings, we were going to perform cluster wise regression to analyze what has been pushing the US Women's team to a consistent high level of performance. This was motivated by the pure disappointment that FIFA does not provide any sort of female analysis.
This goal was unfortunately halted due to lack of data necessary to conduct the steps mentioned above.
What it does
What we ended up creating was a a principal component analysis of the FIFA 18 player ratings to determine desired defensive characteristics. The results were then applied to the Men's World Cup data to illustrate whether defensive characteristic influence wins or losses.
How we built it
Since there were no FIFA Women Player Rating, we decided to use FIFA Men Player Rating to determine characteristics of a good defensive player. We used PCA to reduce the FIFA player ratings to three components, each composed of multiple ratings. We determined the definition of those components and searched for real-world variables in the Men's World Cup data. These real-world measures would provide proxy measures to represent the defensive characteristics. We then visualized the three constructed metrics compared to match results to determine the defensive attributes' influence.
Challenges we ran into
The data did not present a clear or extractable player position. This completely ruined our plan since it was provided in the sample data released earlier this week. The data did not provide possession or results data either. The lack of female player data is disheartening. Due to the above challenges, we had to reformulate our plans multiple times.
Accomplishments that we're proud of
We're proud of the challenges we overcame. We adjusted our ideas, plans, and processes numerous times to adjust with the data and what was deemed possible within he time constraint.
What we learned
We learned the characteristics of a defender. Based upon PCA, we discovered a good defender is made up for three main component tactical skill, athleticism, and aggressive play.
Tactical Skills: Positioning Crossing Curve Long shots
Athleticism: Agility Strength Balance Acceleration Heading accuracy
Aggressive Play: Sliding tackle Long shots Sprint speed Jumping Interceptions Free kick accuracy
What's next for Analysis of Defensive Components
The evidence illustrates that these defensive attributes likely do not drive winning or losing. It would be interesting to continue this research in a manner to see if most outfield players only contribute the bare minimum in defense to conserve energy for the offensive game plan.
Log in or sign up for Devpost to join the conversation.