We're using a custom Python script which was specifically programmed for this giveaway. It uses the YouTube Data API to download a list of all /dbrand and /TLD YouTube subscribers.
Once the full lists have been loaded, the script generates a list of IDs which exist on both lists. This ensures that the potential winner is subscribed to both channels.
From this pool of unique account IDs, a random entry is picked using a randomly generated seed based on the NumPy project. This random pick is the resulting winner.
For every subsequent winner selection, the script will run again. This triggers the YouTube Data API subscriber IDs fetch again. The script will download the newest subscribers and keep going until it reaches the followers which were downloaded during the previous run. What this does is save us from having to download a full list of subscribers every time a winner is selected \u2013 while also ensuring that everyone is included and has an equal chance of winning for each random draw.