deidentify.org tutorial

Step 3 - Data Custodian authorises scripts(s)

Thank you for supplying the raw data and securing the dataset with an encryption password. To ensure the data are non-identifiable to the research team (i.e. that players traces cannot be re-identified via auxiliary attributes, such as which matches the player participated in), we intend to further de-identify the data by running an automated process that extracts just the minimal data needed for the analysis. You (the data custodian) are in full control over which analyses we run, since we cannot decrypt the data without the encryption password.

Analysis 1 - Metadata and Data Sample

This analysis will reveal a sample of each data file to help us understand the data structure, without revealing the entire dataset.

A sample of the first 10 rows and last 10 rows will be extracted for each file.
A summary of values in each column will be extracted (e.g. frequency of each category, mean, standard deviation).

(Complete)

Analysis 2 - Formation Profiles

This analysis will extract team formations in a way that allows studying the team, but not individuals.

Data will be downsampled to one sample every 10 minutes to reduce risk of revealing detailed individual movements.
Team formations will be represented as a “point cloud” that includes non-identifiable dots for the location of each player in the formation, but does not allow tracing the position of individual players over the full course of the match (i.e. even if one knows the player position at a certain time, they will lose track of which player is which whenever two player paths cross over each other).

(For this tutorial, we will authorise the second analysis. Password used to encrypt/decrypt demo data is a2i2.)