Predicting the Outcome of a Cricket Match

Predicting the outcome and events of a cricket match using Hawk Eye data

Rabby, K M Fazly, ID: 10-16482 -2

Abstract: This project aims at predicting the outcomes and as deep as predicting individual events. Previously numerous works have been done on predicting ODI match outcome. One of the most significant ones is the one by Bailey and Clarke that involved data from 2200 ODI matches. They applied Multiple Linear Regression and backward elimination process to identify patterns and variables that could impact the outcome of a match. Their work is mostly based on the Darkwarth-Luis method that could provide only a brief idea on what would be the projected score with a limited number of variables. We, however are working with a lot more complex sets of data that involves manipulation of actual cricket ball trajectory, direction of ball after the batsman plays it and more. We are going to figure out what are a batsman’s weaknesses, what makes a batsman or a bowler nervous, which bowler is most likely to take wicket of a batsman or which batsman is going to make a lot of runs out of which bowler. All of these data will eventually lead us to a more accurate prediction.

Key: hawk eye, ODI prediction, T20 prediction, ball by ball prediction

Introduction:

People have been trying to predict the outcome of cricket matches for a long time. We are yet to see a model that can predict the outcome of a match accurately (the updated Bailey model can predict with around 71% accuracy) and there is none that can predict individual events. So our work is somewhat unique in a sense that we are aiming at predicting major events of a match before it even starts.

To achieve this we needed a huge amount of reliable and precise data. We found the data that http://www.espncricinfo.com/ uses to simulate hawk-eye projection to be almost exactly what we have been looking for which might just hold the key to the solution of our problem.

Methods:

Data Collection Phase 1: We collected json...