2019 ICC Cricket World Cup AI Predictions
Using AI to predict the outcome of the 2019 ICC Cricket World Cup
Cricket has always been a big part of my life since. In particular, the international one-day format is always been my favorite. Even after I moved to North America in my late teens, I would wake up early morning on weekends to watch the matches live.
As I am wrapping up my degree in Master of Data Science from University of British Columbia in Vancouver, BC, Canada, I could think of a better pet-project to put some of my data skills to test than predicting the winner of the 2019 ICC Cricket World Cup.
The primary source of data for this project is ESPNcricinfo.com. The website is truly unparalleled in terms of the wealth of statistics it possesses. Scripts will scrape one-day international matches' and players' details.
The complete data is present in the following files:
match_results.csv: High-level summary of ODI matches include teams, ground, winner and margin
matches_scorecard_details.csv: All player names and URLs, and misc. match information such as attendance
matches_scorecard_player_details.csv: Complete table of player details including style, batting and bowling figures
complete_player_details.csv: Compiled table of ODI teams' players with their attributes and stats
To recreate the data files, run the script with your desired start and end years. For instance:
python scripts/data.py 1971 2019
Sneak Peak at the data
The following bar charts give a sense of the amount of data at hand.
Narrowing in on the top 10 cricketing nations, a look at how well they perform with the bat and the ball.
More to come soon!
Several scikit learn classifiers, including Random Forest, Support Vector Machines and Feed Forward Neural Network, have been used to predict using all the players' batting and bowling statistics. See the perditions here
Python version 3.6.8 and the following python packages:
- pandas (version 0.24.2)
- requests (version 2.21.0)
- bs4 (version 4.7.1)
- re (version 2.2.1)
- sklearn (version 0.20.1)
- matplotlib (version 3.0.3)
- argparse (version 1.1)