discrete project phase 2

Summary

This presentation outlines Phase 2 of a project aimed at identifying optimal passing patterns in football using the overo method. The team divided a dataset of 12,000 football passes into 14 matches, each with 28 player coordinates. They utilized Python for data preparation and the overo technique to compare pattern and target graphs over time. Despite efforts to extract key passing patterns such as 'one-to-one' and 'one-to-four', the project faced challenges, and no matches were found between the pattern and target graphs, resulting in inconclusive outcomes. The analysis emphasizes the complexity of graph alignment in dynamic data.

Takeaways

• 😀 The project focuses on identifying optimal passing patterns in football using the Overo method.
• 😀 Data preparation involved 12,000 rows and 60 columns, including player IDs, timestamps, and X/Y coordinates for 28 players.
• 😀 The dataset was divided into 14 separate matches using Python's Pandas `qcut` function, sorting by time start.
• 😀 Each match was segregated into home and away teams, with 14 players on each team.
• 😀 The Overo method uses pattern graphs and target graphs to match passing patterns across time, ensuring both structural and temporal alignment.
• 😀 The input for the Overo method includes predefined pattern graphs and dynamic target graphs that evolve over time.
• 😀 The code used libraries such as `pandas`, `zipfile`, and `os` to process the data, divide it into matches, and extract the relevant information.
• 😀 The data was split into five time slots per match, each forming a snapshot that was used in the pattern recognition process.
• 😀 Key passing patterns like 'one-to-one' and 'one-to-two-three-four' were examined for their occurrence in the data.
• 😀 Despite attempts to match the predefined passing patterns with the match data, the results were inconclusive, and no isomorphism was found between the pattern and target graphs.

Q & A

What is the main objective of the project discussed in the transcript?

The main objective of the project is to identify optimal passing patterns in football using graph-based methods, specifically focusing on how these patterns evolve over time during matches.

How was the dataset prepared for analysis?

The dataset, containing 12,000 rows and 60 columns, was prepared by dividing it into 14 separate football matches. The rows were sorted in ascending order based on the time start, and then divided into quartiles using the Pandas library in Python.

What does the Overo technique involve in the context of the project?

The Overo technique is a method used to match a predefined pattern graph with a dynamic target graph. It compares both structural and temporal elements, ensuring that the pattern aligns with the changes observed in the target graph over time.

What is the significance of the input data in this project?

The input data consists of football pass data, including timestamps, player IDs, and coordinates for all players on the field. This data is used to create graphs that represent the flow of the game, allowing for the identification of passing patterns.

How were the 14 football matches represented in the dataset?

The 14 football matches were represented as separate datasets, each with a match ID. The matches were further divided into five segments, where each segment represented a snapshot of the match, with each time slot compressed into a single graph.

What challenges were faced during the pattern matching process?

The main challenge was the inability to find an isomorphism between the selected pattern graph and the target graph. This could have been due to errors in the construction of the target graphs or incorrect selection of pattern graphs.

How was the data segmented and processed in the code?

The data was processed by first splitting the dataset into 14 matches using the Pandas `qcut` function. Then, the code created directories for each match and segmented the data into time-based snapshots. Each snapshot was treated as a separate graph to analyze passing patterns.

What role do the home and away teams play in the data segmentation?

The data was divided into home and away teams, with the first 14 players representing the home team and the next 14 players representing the away team. This segmentation helped in organizing the dataset and creating separate target graphs for each team.

Why was the result of the project inconclusive?

The result was inconclusive because, despite various attempts, the pattern matching algorithm did not find any meaningful alignment between the pattern graph and the target graph, leading to zero output. Possible reasons include errors in the construction of target graphs or inappropriate pattern graph selections.

What were some of the key passing patterns explored in the project?

Some of the key passing patterns explored in the project included simple one-to-one passes and more complex sequences such as one-to-two-to-three-to-four passes. These patterns were used to identify frequent and repetitive passing behaviors during matches.