Abstract: We harvest unique "deck-codes" from youtube gameplay videos daily of the popular online trading card game "Hearthstone". We extract and then visualise the decks for users, allowing them to search videos and decks according to the cards being played in them.
Full Description: We are first using "tuber" to access the the YouTube API, extracting video descriptions for videos produced by popular players of the trading card game "Hearthstone". We then extract "deck-codes" from these video descriptions, which are used by Hearthstone to import and share decks between users. We then use Hearthstone specific python libraries, via reticulate, to extract human readable deck composition data (cards within the decks) from the codes, and add them to a SQLite database with RSQLite.
We have then built a catalogue of videos, with metadata attached as well as the codes for importing into the Hearthstone game, links to videos and other online deck building tools. Using DT's HMTL capabilities, we have been able to introduce deck visualisations within the tables, which help users identify the cards within a deck quickly.
The true value of this is that user can filter video content by the decks and cards that are being played within them. It also helps users to find decks to play, and provides a long-term history of the Hearthstone meta and the popularity of different cards and decks. We are also able to provide other visualisations including card synergies and popularities over time.
Table Type: interactive-Shiny
Submission Type: Single Table Example
Table: Wild Deck DB
Code: GitHub - michaelBane/hsDecklistScraper
Languages: Built with R: true. Built with Python: true.
Other packages: DT, reticulate, tuber, RSQLite, shiny