As a Data Scientist and Ultimate Fighting Championship (UFC) fan, there was no way that I wasn't going to build the world's first UFC MMA predictor. I built this web application in late 2018 while waiting for my new job to start at that time. I sourced the data on fighters and UFC fights from public sources and scrapped all the data using a container service that stores it in a sqlite3 database. The predictor is estimated to have an accuracy of 70.4%, better than random guess and picking favourite only!

Per the results from Google Analytics, my web application receives over 200 views per month. Most users are from the U.S and Canada. I have also noticed an increasing amount of citations in academia from my work. Thanks for all your support and glad to have inspired some of you!

• Websitehttps://ufcmmapredictor.herokuapp.com/
• PlatformWeb application deployed on Heroku
• GitHub RepoUFC MMA Predictor
• StackPython, Dash, sqlite3, morph.io (containerized web scrappper)
• Data Analysis & ModellingJupyter Notebook

Media coverage

• Plotly7 of the dash's community latest creation
• Zeming YuMy top 5 visualization tools for data science

Screenshots

Usage of the web application. Select fighters from weight class and then click predict.

Example fight prediction of the headline main event of UFC217 between GSP and Michael Bisping. GSP went on to actually win the fight.