Goal: Predict a player's headshot and kill count in the game CS:GO with machine learning, specifally with regression models.
The dataset that was used to train the models cannot be disclosed. However, it involves historical data for ALL players who are involved in competitve gaming. The data is also stored in a MySQL database that is operated on AWS.
The model that was trained and exported for making future predictions was an XGBoost Regressor.
R2 Score
- Kills: 0.7199
- Headshots: 0.645
Mean Absolute Error
- Kills: 5.658
- Headshots: 3.483
Root Mean Square Error
- Kills: 7.3899
- Headshots: 4.552
The R2 scores are moderate indicating a decent fit. When backtracking the results, on average, the unders have a higher hit rate. The overall performance fluctuates and hovers around 52% hit rate.
I cannot disclose anymore on this project, but the code and modeling can be found in this repository. Feel free to also contact me using the link below.