To analyze and predict which grappling techniques and positions are more likely to result in control and submission in martial arts, using data-driven methods.

This study utilizes the "Grappling Techniques" dataset from Kaggle, compiled by Luca Besso. It includes techniques from Brazilian Jiu-Jitsu, Judo, and Wrestling. The dataset is under the Apache 2.0 License and is accessible here.

The dataset was preprocessed for missing values, and an analysis was conducted to determine the distribution of techniques and positions. Our analysis incorporates advanced statistical techniques and comparative studies across various martial arts styles.

We created bar charts to depict the frequency of dominant versus defensive positions. Additional visualizations compare the effectiveness of techniques across Brazilian Jiu-Jitsu, Judo, and Wrestling, highlighting style-specific trends.

A RandomForestClassifier is employed to predict the effectiveness of positions in grappling. The model's performance was evaluated using metrics such as accuracy, precision, recall, and cross-validation. Feature engineering was applied to enhance predictive capabilities.

The predictions from our model, coupled with extended analysis, offer insights that may indicate future trends in martial arts techniques.

• Install Python and the required libraries (Pandas, Matplotlib, scikit-learn).
• Execute analysis.py for initial analysis and data visualization.
• Refer to extended_analysis.py for advanced analysis and visualizations.

• analysis.py: Contains initial analysis, machine learning model training, and advanced statistical analysis with visualizations.
• dataset.csv: The dataset file used for the project (not included in this repository).

Detailed are the model's accuracy, precision, recall, and a classification report. Feature importance is evaluated to understand influential factors better. Key findings from our extended analysis are also presented.

Contributions are welcome. Fork the repository, commit your changes, and create a pull request. See our contribution guidelines for more details.

This project uses the "Grappling Techniques" dataset under the Apache 2.0 License, as provided by Luca Besso on Kaggle.

For inquiries or collaboration, connect with Omeed Tavakoli on LinkedIn.

Special thanks to the martial arts community, Kaggle, and Luca Besso for providing the dataset.