WotHub is a data mining project for World of Tanks in which we parse data packets from game replays, preprocess data and make players position predictions based on statistical approach.
Only Windows OS is supported.
If you want to check out web version, jump to "Deployment" section.
- Make sure to have Python 3+ along with all necessary libraries from
requirements.txt. - For a parsing part we use wotreplay-parser created by Jan Temmerman. Download parser binaries and extract in the repo directory.
- Download replays and place them in
/replaysfolder. The more, the better.
- run
parser.py
Parser reads replay files from /replays folder. Parsed data will appear in /raw_data folder in .json format for each replay.
- run
preprocessor.py
It reads json files from /raw_data folder. Preprocessed data will appear in /pro_data folder in .csv format, accumulated for each map.
Visualization consists in creating position density plot where high values represent most frequent players positions.
Note: Heroku does not store csv files, so we load data from AWS S3 Server, instead of
/pro_datafolder. So, if you run visualization on localhost, you can use/pro_datafolder.
We use streamlit for convenient manipulation.
streamlit run .\web_visual.pyruns visualization on localhost
Example:
We use heroku for further project deployment. Here is a link to web app - WotHub (depricated).
- Update maps.
- Change density function parameters for smoother plot.
- Add Unix support.
- Archive repo.