A Novel Graph Based Approach to Predict Man of the Match for Cricket

The advent of computational intelligence has only accelerated the motive to automate manual processes. In sports, the award of the best player is questionable in many cases and requires the opinion of multiple distinguished experts in the sport. This often results in an unfair judgement of the situation, given the time constraint for making this decision, and the dependence on pure human skill and reasoning. This calls for a reliable and fair system to fairly award the man of the match, taking into account the numerous variables. Though there are machine learning based approaches trying to model the game for predicting win/loss outcome, it is hard to model the “man of the match” in a cricket game using the same approach. We propose a novel graph based approach to award the “man of the match” to a player for a cricket game. We model the game as a directed graph with each player as a node and an interaction between two players as an edge. We use ball by ball delivery details to construct this graph with a heuristics to calculate the edge weights and compute node centrality of every player using the Personalized Pagerank algorithm to find the most central player in the game. A high centrality indicates a good overall performance of the player. Comparison with existing game data showed encouraging results.