By Topic

Evaluation and Extension to the Duckworth Lewis Method: A Dual Application of Data Mining Techniques

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Viraj Phanse ; Dept. of Comput. Sci., Univ. of California, Los Angeles (UCLA), Los Angeles, CA, USA ; Sourabh Deorah

This paper deals with the evaluation of the Duckworth Lewis method, identifying its limitation, and devising a modification to address these limitations. The Duckworth Lewis method, or D/L method, was created by Frank Duckworth and Tony Lewis. The International Cricket Council (ICC) adopted D/L method in 1999 to address the issue of delayed one-day cricket matches due to interruptions such as inclement weather conditions, poor light and floodlight failures, and crowd problems. We have attempted to identify the shortcomings in the existing Duckworth Lewis method using data mining algorithms such as Random Forests and C4.5. We have also shown that the p-values and other data mining techniques serve a dual purpose of not only evaluating whether systems such as D/L method have been exploited by taking advantage of their properties such as simplicity, but also devising alternate and robust approaches (or models). In the first part of this project, we have analyzed fifty one-day international (ODI) cricket matches, in which the Duckworth Lewis method has been applied, using tools such as WEKA and Microsoft Excel. We have observed that the Duckworth Lewis method has some limitations. As a result, using data mining methods we have shown that the Duckworth Lewis system has proven over time to be biased towards the team batting first and the team winning the toss -- a toss refers to the coin-flip at the beginning of the match used to decide who bats or fields. Bias in the context of the report is defined as taking advantage of the properties of systems such as the Duckworth Lewis method. We also seek to show that such an "exploitation" of the system permits prediction of the match winner with outcomes that are better than just chance. Using the analyses described above, we propose a modification to the existing Duckworth Lewis method to reduce this bias by considering the "venue" of the game as an additional resource along with the two existing resources-overs and wickets - - o predict the target score. We have done a basic evaluation of the reduction in bias due to the proposed changes. The modification has helped not only to reduce the bias but also to alleviate the impact of factors such as toss and team batting first in predicting the target scores in limited-overs cricket matches.

Published in:

2011 IEEE 11th International Conference on Data Mining Workshops

Date of Conference:

11-11 Dec. 2011