Predict DotA2 players' possibility of quitting using logistic regression
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Final Project for CMU 19704

In this project, we analyze DotA players' data downloaded from the OpenDotA website, where we can get access to over a million DotA players data ever since their first game. Our goal is to build a logistic regression model to predict when a player will decide to quit the game. In total we downloaded 35352 players’ data ( Each player has one json file which contains all games that player was ever involved. Each game has a data point in the json file, formatted as follows:

Using the python script, we calculated independent variables:

We applied 20/60/20 rule to our data analysis and finally got a model with 81% prediction accuracy. More details of the analysis can be found in the report: