UFC Prediction

In this project, we scrape data from ufcstats.com and apply machine learning techniques to this data to make UFC fight predictions (winner and method). The predictor is available here.

For the Developer

Make sure Python 3 is installed on your computer. Clone this repository. Open a terminal and cd to the location you saved it.

The dataset is already included in the repository (last updated April 9, 2022) so there is no need to scrape the stats and build the dataframe from scratch, but you can do so from the jupyter notebooks in src/models/buildingMLModel/notebooks/ if desired. In particular, you can run all cells in UFC_data_scraping.ipynb to scrape the entire dataset from the source ufcstats.com, and run cells in Building_ufc_fights.ipynb to compile the dataframe we use to build our ML model (though this would take weeks on a laptop). To scrape and stylize fighter pictures used on the website, you can run UFC_picture_scraping.ipynb and Pytorch Convolutional Neural Network Style Transfer.ipynb. The rest of the notebooks are dedicated to building and testing machine learning models for fight and method prediction.

After every UFC event, the dataset and model will need to be rebuilt. Follow these instructions to rebuild both:

1. First we will scrape ufcstats.com and update the dataset. Open a terminal in the UFC_Prediction_2022 directory and run

cd src

followed by

python3 update_data_csvs_and_jsons.py

(this will take awhile to run... about 20 minutes per event since the last update)

2. Now, type in your terminal:

python3 update_and_rebuild_model.py

and press enter. This will run more quickly. This rebuilds the machine learning model to incorporate the updated data. This will send the updated coefficients to the files theta.json and intercept.json which are used on the website to implement the current build. It also scrapes vegas odds for upcoming events from here, makes predictions for upcoming events, and updates the json files used to populate tables for the website.

3. Now, to make sure the current build is working, cd to the main directory UFC_Prediction_2022 run the following from the terminal

python3 -m http.server 8000

Now go to chrome and type localhost:8000 into the address bar. This opens the updated version of the website.