Predicting the Outcome of ODI Cricket Matches: A Team Composition Based Approach Madan Gopal Jhanwar, Vikram Pudi Center for Data Engineering Kohli Center for Intelligent Systems Internation Institute of Information Technology - Hyderabad, India September 19, 2016 Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 1 / 19
Overview Introduction 1 Related Work 2 Our Approach 3 Experiments 4 Results 5 Conclusion 6 Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 2 / 19
Introduction With the advent of statistical modeling in sports, predicting the outcome of a game has been established as a fundamental problem. Cricket is one of the most popular team games in the world. Various natural factors affecting the game, enormous media coverage, and a huge betting market have given strong incentives to model the game from various perspectives. However, the complex rules governing the game, the ability of players and their performances on a given day, and various other natural parameters play an integral role in affecting the final outcome of a cricket match. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 3 / 19
Introduction With this article, we embark on predicting the outcome of a One Day International (ODI) cricket match using a supervised learning approach from a team composition perspective. Our work suggests that the relative team strength between the competing teams forms a distinctive feature for predicting the winner. Modeling the team strength boils down to modeling individual players batting and bowling performances, forming the basis of our approach. We use career statistics as well as the recent performances of a player to model him. Player independent factors have also been considered in order to predict the outcome of a match. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 4 / 19
Related Work Duckworth and Lewis proposed a solution, called the D/L method [1], to reset targets in rain interrupted matches which was adopted by the International Cricket Council (ICC) in 1998. The use of Duckworth-Lewis resources to assess players performances has been studied in [1], [2] and [3]. The methods of graphical representation to compare players are presented in [4], [5], and [6]. [7] considers the strength of opponent team, along with other factors, in modeling the performance of batsmen and bowlers. However, like in any sport, winning is the ultimate goal in cricket. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 5 / 19
Related Work [8] takes into account various factors affecting the game including home team advantage, day/night effect and toss, etc., and uses the Bayesian classifier to predict the outcome of the match. [9] uses a combination of linear regression and nearest-neighbor clustering algorithms to predict the outcome of a match. They take into account both historical data as well as instantaneous state of a match while the game is still in progress. [10] studied the role of multiple factors including home field advantage, toss, match type (day or day and night), competing teams, venue familiarity, and season, etc., and applied Support Vector Machines(SVM) and Naive Bayes Classifiers for predicting the winner of a match. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 6 / 19
Novelity In this paper, we embark upon a very critical aspect that the team composition changes over time, which has not been studied yet. A team is comprised of 11 players, and these 11 players are replaced over time. A team changes its composition depending upon the match conditions, venue, opponent team, etc. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 7 / 19
Novelity 3.0 2.5 No of player changes per match 2.0 1.5 1.0 0.5 0.0 a a d d a a n s h i c n n d i k a e s a l i a a n t i e r n s d d t r A f g l a l I a i n s n L k a u e a I l h E Z r i P t g A t S s n u w e a o W S e B N Countries Figure 1 shows that on average at least 2 players change per match for each team. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 8 / 19
Our Approach To predict the winner of ODI cricket matches, we propose a novel dynamic approach to reflect the changes in player combinations. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 9 / 19
Model Batsman Feature Description Matches Played #Matches played by the player Batting Innings #Matches in which the player batted Batting Average #Runs scored divided by the #times the player got out Num Centuries #Times the player scored ≥ 100 runs in a match Num Fifties #Times the player scored ≥ 50 but less than 100 runs in a match We use a combination of these features to estimate the Career Score of a batsman. We also consider the player’s recent performances to analyze his prevailing form. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 10 / 19
Model Bowler Feature Description Matches Played #Matches played by the player Bowling Innings #Matches in which the player bowled Wkts Taken #Wickets taken by the player FWkts Hauls #Times the player has taken ≥ 5 wickets in a match Bowling Average #Runs conceded by the player per wicket taken Bowling Economy Average #runs conceded by the player per over bowled We use a combination of these features to estimate the Career Score of a bowler. Due to the lack of the data, we could not analyze the recent performances of a bowler. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 11 / 19
Modeling Teams We first sum over the inidividual player potentials to calculate the team batting and bowling potentials. We then use the calculated team batting and bowling potentials to estimate the relative strength of one team against the other. Our algorithm follows the fundamental aspect of the game strategy where the batsmen of one team work against the bowlers of the other team and vice-versa. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 12 / 19
Experiments Dataset The dataset includes all the ODI matches played between 2010 and 2014. We have restricted our study to only top 9 ODI-playing teams. All the matches that ended up without a result have been removed. The training dataset contains all the matches played during the years 2010 to 2013, and the test dataset contains all the matches played in the year 2014. Learning Weights To assign weights to various features in multiple algorithms, we have used the 5-match ODI series played between India and Sri Lanka in July, 2012. A series of consecutive matches was deliberately chosen to study the impact of the recent scores of a batsman on his upcoming performances. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 13 / 19
Results 0.72 0.70 0.68 0.66 Accuracy 0.64 0.62 0.60 0.58 M t n s N s V e o e N S i e K r s r o r s T e F r n g o m e i s o R i d c n c e t i D a s R i g o L Algorithm Accuracy of the different classifiers used. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 14 / 19
Results The only obstacle we faced while evaluating our approach is the inability to compare against previous models due to the different underlying datasets used. However, we compared our model with two other baseline models. Table : Comparing our kNN based model with other baseline models Model Accuracy Toss Wins 0.56 Rel. Strength Wins 0.63 Our Model 0.71 The superiority of our model against the others proves the significance of the combination of various features used. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 15 / 19
Conclusions The paper addresses the problem of predicting the outcome of an ODI cricket match using the statistics of 366 matches. The novelty of our approach lies in addressing the problem as a dynamic one, and using the participating players as the key feature in predicting the winner of the match. We observe that simple features can yield very promising results. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 16 / 19
References Duckworth, Frank C., and Anthony J. Lewis. ”A fair method for resetting the target in interrupted one-day cricket matches.” Journal of the Operational Research Society 49.3 (1998): 220-227. Beaudoin, David, and Tim B. Swartz. ”The best batsmen and bowlers in one-day cricket.” South African Statistical Journal 37.2 (2003): 203. Lewis, A. J. ”Towards fairer measures of player performance in one-day cricket.” Journal of the Operational Research Society 56.7 (2005): 804-815. Kimber, Alan. ”A graphical display for comparing bowlers in cricket.” Teaching Statistics 15.3 (1993): 84-86. Barr, G. D. I., and B. S. Kantor. ”A criterion for comparing and selecting batsmen in limited overs cricket.” Journal of the Operational Research Society 55.12 (2004): 1266-1274. Madan Gopal Jhanwar, Vikram Pudi (IIIT-H) ODI Cricket Matches September 19, 2016 17 / 19
Recommend
More recommend
