Overwatch player analysis for PSU Stat 184 final project. The goal of this project is to be able to determine player rankings from player data.
Run Rscript overwatchAnalysis.r to run the analysis. Only the analysis runs by default Current run options:
- --install
- If equal to true will install all required packages from default CRAN mirror
- --tags
- If equal to true will scrape battle tags from first 950 pages of OW PC leaderboards
- --pro
- If euqal to true will scrape profile data from battle tags. This is an extremely slow process and scrapes approximately 9500 records be default.
- --a
- Default value is true and performs the data analysis including training model and creating graphs.
- In the future will have separate options for different analysis functions.
You must run install.packages('optparser') if you wish to run with options.
The machine learning model can currently only predict within 400 SR on average. I believe that if I scrape a lot more records this will improve, since I saw an improvement using 8,000 training rows versus 6,500 rows.
One of the reasons I only had 10k records was the insanely inefficient scraping that took palce. I hope to find a more efficient method to then scrape everything and run it through the model again.