2019-clashroyale-analytics

Analysis of Clash Royale matchs to generate useful statistics. (Simón Sepúlveda, Matías Saavedra, Daniel Guzmán)

Overview

In this project we work with a Clash Royale Matches dataset which contains about 700k records of matches. The main objetive is to generate useful statics about the cards and decks performance like the most winning/losing cards/decks, the most used/unused cards, etc. With this information we generate graphics and tables for a good results visualization.

Data

We work with a Clash Royale Matches dataset which contains about 700k records of matches in .json format. The dataset starts at 2017 - 07 - 12. The dataset can be found in Kaggle at https://www.kaggle.com/s1m0n38/clash-royale-matches-dataset .

This dataset was chosen because 66.666% of the team are active players and have a lot of fun and also frustration when losing with Clash Royale. The 33.333% remaining are afraid of the fame and sex appeal the game generates on players.

Each tuple of the dataset looks like:

{'players': {'right': {'deck': [['Mega Minion', '9'], ['Electro Wizard', '3'], ['Arrows', '11'], ['Lightning', '5'], ['Tombstone', '9'], ['The Log', '2'], ['Giant', '9'], ['Bowler', '5']], 'trophy': '4258', 'clan': 'TwoFiveOne', 'name': 'gpa raid'}, 'left': {'deck': [['Fireball', '9'], ['Archers', '12'], ['Goblins', '12'], ['Minions', '11'], ['Bomber', '12'], ['The Log', '2'], ['Barbarians', '12'], ['Royal Giant', '13']], 'trophy': '4325', 'clan': 'battusai', 'name': 'Supr4'}}, 'type': 'ladder', 'result': ['2', '0'], 'time': '2017-07-12'}

Each tuple contains:

• players: Can be left or right player.
  • left/right: Only for players reference. Contains:
    • deck: The player deck with their eight cards played and their level.
    • trophy: The player ladder trophies.
    • clan: The player clan.
    • name: The player name.
• type: The match type. Can be ladder, challenge or tournament.
• result: The result of the match in the format [left,right]. It can be all combinations beetween 3-0 and 0-3.
• time: The date when match was played.

Also we work with a dataset that contains the attributes of each cards as the Elixir cost, the card type, the card damage, etc. This dataset is used only for add useful information but not for querying. This dataset can be found it at https://www.kaggle.com/swappyk/clash-royale-dataset .

Methods

MongoDB is used to process the queries made because of its compatibility with the data's format and its already available by the course's faculty members for which this project is developed.

For each desired statistic a pipeline is designed, which is presented on Results with a description of the resulting table (collection).

Python is used to make the visual representation of the data, because of its ease of use for the team and its variety of methods to further process the data (only for ordering, no further data is collected here). For this,we use the matplotlib library for plotting, pandas and json for loading the exported results and numpy for data processing. This last library is needed because the results are not always congruent between queries: for example on certain matches or kind of cards not all levels are on the databse because of its rarity, so we need to complete or truncate certain results.

Results

We show some pipelines of the MongoDB's queries made along with some interesting graphics made on Python with the collections obtained. The whole set of queries can be found on the repository on the queries.txt file.

Players with the most played games on ladder type matches by descending order:

Decks played by a particular player

Results obtained by a player playing with a specific deck

Times a card was played on a won match grouped by level

For the graphics made, we show the winrate by level of specific cards, the use of specific cards by its level across all games, and the use histogram of all cards on won, lost and drawn matches.

Cards histogram:

Winrate:

Cards use by level:

Conclusion

We developed a way to obtain useful statistics about the Clash Royale mobile game using MongoDB by generating queries and defining functions to obtain some interesting data about the outcome of games and players and use this data to represent it in a more visually appealing way via Python to give it (potentially) to players and enhance their gameplay. Taking this into account, we consider the objective of the project as achieved but not easily achieved: the main difficulty was in fact using MongoDB due to its complex syntax which wasn't very clear for the team in many cases.

Respect to the database, we found that higher level cards were played much more times than low level cards, resulting in not having much info about the latter so the graphics made could not have been very informative. An even bigger database would help for this purpose, but for effects of the course it was sufficient.

Finally, the project development was considered as a good instance to further learn about the tools used on the course, because it pushed us much farther than the laboratories did, having acquired a vast knowledge about MongoDB , hoping that it would come as useful on our carees through our lifes.

Appendix

See queries.txt on this repository to see all the queries made for the development of this project.