Abstract:
In Sri Lanka, cricket is the most popular game and has become an integral part of the culture.
Various factors come into play while selecting a team. A human selection committee invariably
suffers from the shortcomings of unfair or biased judgment and human errors. A system is thus
required which can effectively take into account all factors involved and give the optimal team,
without human interference. This study is to develop a scientific method for selecting the best
batsmen for a particular 50 over match. The study was carried out under the Sri Lankan context.
Critical factor of this study does not use traditional batting measurements, instead focus on a
measurement, which reflects the performance of a 50 over cricket match. Moreover
measurements were developed to assess players‘ performance against different oppositions,
grounds and recent foam. The graphical analysis showed that Sri Lanka as a team and
individual players had not performed consistently against different oppositions or at different
grounds. One major finding is that not every player‘s performance is low or high at a given
ground or opposition. This revealed that player performance depends on the opposition and
ground. The performance of cricketers depends on their talent and their experience, which in
turn results in an obvious hierarchical structure, where the performances are clustered within
the cricketer. This prompted in adopting Generalized Estimating Equations (GEE) as it is one of
the methods based on the model performance measurement was calculated and then ranked the
players to pick the best set of players for the match. Genetic Algorithm, inspired by biological
evolution, was used to perform a user-defined task with no assumptions to be satisfied. This
machine learning technique used to optimize the set of batsmen using a defined objective
function. Opposition, ground and recent performance are taken as input variables in optimising
the fitness value to arrive at an optimal set of batsman. Both the methods demonstrate valid
results in achieving the desired objective when compared with the original match results.
Therefore, these methods will provide a basis for selectors to overcome many issues.
