The game of Cricket, especially its limited overs version, is full of glorious uncertainties, which is what makes it so much fun to watch and follow. Because there are so many factors that determine the outcome of a match – team compositions, weather, pitch conditions, etc, it may appear on the surface that predicting the outcome of cricket matches might be hard. However, in this article we present an approach that has been used to predict the outcome of Cricket matches with at least two-thirds accuracy in the ongoing World Cup.
The Methodology
Our current approach is based solely on the team compositions. First, our method calculates a score for each player of the team based on their past performances. This score is calculated by looking at the one-on-one match-up between the batsmen and the bowlers in their previously played matches (more on this later). For a given match, we sum up the total score of the top eleven players in each team, and the resulting win probability of a team is proportional to its total score. The team with the higher win probability is the predicted winner of that match.
As for computing the player scores, the objective is two fold. First, it should be able to combine both the batting and the bowling statistics of a player into one score that can be used in our prediction algorithm. The second is that it should be able to adjust the score of a player based on the quality of the opposition that player faced. That is, a player’s score should not be higher just because he boosted his statistics by playing against weaker teams. To ensure that both these objectives are met, we employ a “page-rank” type of an algorithm for computing the player scores, that uses the batsman versus bowler match-up data from previously played matches. More details about this methodology can be found on the Cricmetric website. For computing player scores for predicting the outcome of Cricket World Cup matches, we used the data of the ODIs played since April 1, 2004.
Prediction accuracy
For the World T20 tournament in 2014, our prediction method correctly predicted 15 out of 23 matches of the league stage of the tournament, for a success rate of 65%. It did quite well in the league stage of the tournament, correctly predicting the outcome of 15 out of 20 matches, but got all the three predictions wrong for the knock-out stage of the tournament. For the ongoing Cricket ODI World Cup 2015, so far we have correctly predicted 21 out of the first 31 matches of the tournament.
A complete list of the previous and the future predictions (as of March 8, 2015) for the Cricket World Cup are given in the table below.
Date | Team 1 | Team 2 | Cricmetric’s prediction | Actual result |
---|---|---|---|---|
Feb 14, 2015 | New Zealand | Sri Lanka | New Zealand (70%) | New Zealand |
Feb 14, 2015 | Australia | England | Australia (56%) | Australia |
Feb 15, 2015 | South Africa | Zimbabwe | South Africa (76%) | South Africa |
Feb 15, 2015 | India | Pakistan | India (59%) | India |
Feb 16, 2015 | West Indies | Ireland | West Indies (56%) | Ireland |
Feb 17, 2015 | Scotland | New Zealand | New Zealand (78%) | New Zealand |
Feb 18, 2015 | Bangladesh | Afghanistan | Afghanistan (80%) | Bangladesh |
Feb 19, 2015 | U.A.E. | Zimbabwe | Zimbabwe (70%) | Zimbabwe |
Feb 20, 2015 | England | New Zealand | New Zealand (59%) | New Zealand |
Feb 21, 2015 | West Indies | Pakistan | Pakistan (54%) | West Indies |
Feb 21, 2015 | Australia | Bangladesh | Australia (64%) | ABANDONED |
Feb 22, 2015 | Afghanistan | Sri Lanka | Sri Lanka (69%) | Sri Lanka |
Feb 22, 2015 | India | South Africa | South Africa (62%) | India |
Feb 23, 2015 | England | Scotland | England (73%) | England |
Feb 24, 2015 | West Indies | Zimbabwe | Zimbabwe (53%) | West Indies |
Feb 25, 2015 | U.A.E. | Ireland | Ireland (60%) | Ireland |
Feb 26, 2015 | Scotland | Afghanistan | Afghanistan (62%) | Afghanistan |
Feb 26, 2015 | Sri Lanka | Bangladesh | Sri Lanka (70%) | Sri Lanka |
Feb 27, 2015 | South Africa | West Indies | South Africa (70%) | South Africa |
Feb 28, 2015 | Australia | New Zealand | Australia (52%) | New Zealand |
Feb 28, 2015 | U.A.E. | India | India (78%) | India |
Mar 01, 2015 | England | Sri Lanka | Sri Lanka (59%) | Sri Lanka |
Mar 01, 2015 | Pakistan | Zimbabwe | Zimbabwe (54%) | Pakistan |
Mar 03, 2015 | South Africa | Ireland | South Africa (81%) | South Africa |
Mar 04, 2015 | Pakistan | U.A.E. | Pakistan (68%) | Pakistan |
Mar 04, 2015 | Australia | Afghanistan | Australia (62%) | Australia |
Mar 05, 2015 | Scotland | Bangladesh | Bangladesh (58%) | Bangladesh |
Mar 06, 2015 | West Indies | India | India (62%) | India |
Mar 07, 2015 | Pakistan | South Africa | South Africa (71%) | Pakistan |
Mar 07, 2015 | Ireland | Zimbabwe | Zimbabwe (63%) | Ireland |
Mar 08, 2015 | Afghanistan | New Zealand | New Zealand (60%) | New Zealand |
Mar 08, 2015 | Australia | Sri Lanka | Sri Lanka (57%) | Australia |
Mar 09, 2015 | England | Bangladesh | England (63%) | Not started |
Mar 10, 2015 | India | Ireland | India (69%) | Not started |
Mar 11, 2015 | Scotland | Sri Lanka | Sri Lanka (78%) | Not started |
Mar 12, 2015 | South Africa | U.A.E. | South Africa (84%) | Not started |
Mar 13, 2015 | New Zealand | Bangladesh | New Zealand (63%) | Not started |
Mar 13, 2015 | Afghanistan | England | England (59%) | Not started |
Mar 14, 2015 | India | Zimbabwe | India (59%) | Not started |
Mar 14, 2015 | Australia | Scotland | Australia (73%) | Not started |
Mar 15, 2015 | U.A.E. | West Indies | West Indies (69%) | Not started |
Mar 15, 2015 | Ireland | Pakistan | Pakistan (62%) | Not started |