A Generic Approach for Team

Selection in Multi–player
Games using Genetic
Algorithm
Arvind kumar Stayawan patel Kapil bahrti Amit kumar
0010011 0010030 0010015 0010005
dept. (Master computer Application)

ABSTRACT
The best team selection is always the requirement of management in different
domains and in different organizations including government, project, industry,
business and sports. The traditional team selection process is really lengthy,
awkward and unclear due to manual process and personal judgments; which may
lead to a disaster. These conflicting constraints and personal judgments can be
translated into software for better and quick solution.This paper presents a solution
to this problem with the help of genetic algorithm to find the optimal solution for the
problem of cricket team selection and formation.
Our approach is the combination of the existing quantitative approaches with some
new extensions such as attributes regarding personal performances, team
performance and the combination of players. Secondly our method is just not
specific for cricket team but it is converted to generic model for other multiplayer
games. We propose an adaptation of island genetic algorithm to optimize the
selection of multiplayer sports team having multiple conflicting constraints with
mixed crossover where the fitness of common solution is used to drive the selection.

General Terms
Genetic Algorithm, Sports, Optimization.
Keywords
Player Selection, Genetic Computing, Multi-Player Sports, Team Selection, Cricket.
1.INTRODUCTION
Process of team selection in multi-player sports is a complex decision due to many
factors. The performance of the team depends upon the combination of team
players selected with respect to the performance, event, opponent and the playing
conditions. The confidence of the team is dependent on various factors which may
include the recent performance of the players, the quality and fairness of the game,
satisfaction of selection board or committee including coaches and captain;
moreover the spectators. Aforesaid factors significantly improve the chances of win
with an optimal team selection. Team Selection in multiplayer sports such as
football, hockey and cricket is a daunting task. Composition of the team depends on
many factors which include the player‟s individual performance, combination of
players, physical fitness, playing conditions and above all, the confidence of the
selection committee on the players. Selection of players in a team is always a
difficult decision making task and it has many dimensions. There are finite set of
variables with many constraints. The best team selection is a real world problem for
sports management and associations especially in multi-player sports. Team
selection problem may rise in various domains i.e., software project team, in HRM for
promotion, software testing criterion and bench mark selection etc.The manual team
selection process may have many drawbacks such as personal liking and disliking,
personal grudges between team selection committee and players and external
pressures (social, political etc.). Poor team selection lead to failure and the selection
committee and sports organizing authorities will become accountable to the people
who love sports. It also effects the loyalty of players due to non-optimal and
substandard selection of combination; causing global damages for the side. The
team selection procedure also takes and ample amount of time and chances of
defacement exists. This may results in a bad performance in the game and the
morale of the whole team shattered. To avoid this vague practice of team selection a
scientific and more effective procedure can give an optimistic result in a lesser
amount of time [1]. We have proposed a software based solution for the optimal
team combination by using evolutionary programming technique called genetic
algorithm [2]. We have applied our proposed GA for the selection of cricket team as
it is one of the most popular games in South-East Asia and specifically in India and
Pakistan. Cricket team is comprised of 11 playerscategorized as bowlers, batsman
and wicket keeper. However there are other categories as well such as all-rounder (a
bowler who can bat or vice versa). Similarly batsman can be categorized as (top
order or middle order). Cricket has different formats testcricket (5 day match), one
day international (50 over match) and T-20 (20 over match). Team selections are
based on the format of the cricket a player may be rated well for test cricket may
not be considered for one day international or vice versa. Very few research papers
found in the literature regarding GA in team selection, however [4] and[7] presented
GA based solution for optimal cricket team selection. The next section gives a brief
introduction of genetic algorithm and the problem definition. Next section presents
the flow chart of applying genetic algorithm to team selection process. Finally the
players data is processed using the propose model and results are displayed.

2.GENETIC ALGORITHM IN A NUT SHELL

Evolutionary algorithms such as geneticalgorithm may be used for optimization of
real world problems. Genetic Algorithms (GAs) [2], [3],[8] and[9] are adaptive
heuristic search algorithm designed to simulate processes in natural system
necessary for evolution where we have a huge search space. GAs compared to the
traditional methods, provides an alternative methods to solving problem by
outperforming the classical methods in many problems domain [6].Team selection is
also an optimization problem as the selection committee looks for the best
combination available with given constraints. GA, as discussed earlier considered
being the best tool for optimization in such problems [5]. GA with its power of
computation can solve this problem in a much desirable and efficient manner which
can make the job of selection committee much easier. Let‟s consider a pool of 100
players and the task of the selection committee to select 15 players for the
competition. There may be many combinations of best team players, provided the
realities and constraints. Selecting the best team (optimal) combination is the main
objective and we also call this the fittest combination. The procedure of finding the
fittest combination is a time consuming task and the resultant or the survivor among
the fit candidates leads. In other words, we have to find the “Survivor of the fittest”.
2.1 Working of GA in Context
GA works by initializing the population (Chromosomes) as elaborated below (see
Figure1).Secondly, Selection is made randomly (Rolette wheel selection
algorithmmay be used to avoid the influence of initial population), using the fitness
function. Now the process of Crossover is to be made and at last Mutation
(Optional). This will be repeated in a cycle till the user intervene or fittest
combination found.

2.2 Limitation and Way Out of GA

Besides many strong problem solving strengths, GAs are superseded by any other
modal specific algorithms, where universe of discourse is not complex, so, this is in
fact very true that it has boundaries as well and eventually not successful in every
situation [6]. One other issue is their time complexity which makes it not suitable for
the real time applications. But to override this problem, we may pose the
implementation of this Modality in parallel computation; toreduce the time
complexity limitations and attain success.GA launched with all candidate solutions.
A Model of modal is required to be formulated so the fitness or the degree of
happiness can be calculated for feasible solution. Feasible solutions in theuniverse
of discourse are collected based on following major steps.
a. Evaluation or the degree of fitness is tagged up to respective candidate after
calculation. Initially, various pairs of candidate solutions are selected with higher
degree of fitness. These candidate solutions are then dispatched to the mating pool
for Crossover.
b. Mutation is performed if necessary
Step–„a‟& step–„b‟continue-on in iteration till a suitable, necessary and sufficient
solution with a particular number of generations have attained. That is again based
on the degree of happiness for the specific situation. Evaluation and cost of fitness
function may be matter of concern and settlement standard is the modality when
better candidates are not improving well; after various iteration.

Pr( kj)= fitness(ki) / Σfitness(kj) .........................1.

Where,
Pr(ik)= fitness of current kthchromosome
n = number of genes or size of chromosome (i.e., total number of players)
i = number of currently available players
j = Number of all available players
An approach to designate a fair chance to the best candidates is mentioned in
equation
(i). Better generations are replaced with if parents are not better.

3.MODULAR APPROACH FOR TEAM

SELECTION USING
GAEventually, a modular approach as shown in is proposed to adopt appropriate
optimization technique considering different possibilities in aforesaid team selection
problem [8].GA is allowed to evolve until best chromosome is not obtained. Stopping
criteria of GA is discussed in next section.
3.1 Generate Initial Population
In the proposed method we have made initial population of 10 chromosomes,
initialized randomly. Our chromosomes contains randomly generated player IDs
from the available set of players.
3.2 ApplyingGA
After generation of initial population, Crossover and Mutation operations are applied.
Crossover is applied on two randomly selected chromosomes from our population.
Similarly Mutation is applied on a randomly selected chromosome with a mutation
rate of 0.5
3.3 Calculate Fitness Values of
Chromosome
sIn the next step fitness values of chromosomes are calculated using equation
(ii) and equation (iii).
3.4 Regenerate Population
New population is generated using best 5 chromosomes from population based on
fitness values calculated in above step. Other five chromosomes are newly
generated randomly to complete set of 10 chromosomes.

4.GAFOR PLAYER SELECTION

Team is a combination of different types of players whichshows a sample
chromosome for a team of 16 players with 4 types of player categories. The
description of squad is further elaborated .
W = number of type-1 players. (Gene1 to Gene 8) (W = 8 in our case)
X = number of type-2 players. (Gene9)(X= 1 in our case)
Y = number of type-3 players. (Gene 10 to Gene 14)(Y = 5 in our case)
Z = number of type-4 players. (Gene 15 to Gene 16)(Z = 2 in our case)

5.Results & Discussion

There may be various combinations exist; pitching an optimal solution is inevitable.
Therefore, genetic algorithm like heuristic algorithm may be tailored to solve this
problem. We have applied proposed method on currently available players of
Pakistancricket team, shown above in Table 3. Based on our proposed fitness
function of GA, we obtained fitness values of all players Based on these
fitnessvalues following team combination is obtained shown below in Table 4and
(see Figure4) thatshows best chromosome evolved after 2257 generations of GA.
Our obtained fitness value for this chromosome is 0.9544220740625.

6.CONCLUSION ANDFUTURE WORK

A generic approach for multi player team selection has been developed using
genetic algorithm. This automated process will select team on the basis of the
previous outcomes i.e. either win or lose. This algorithm also takes care of the
recent performance of the players by considering last M matches. We have applied
this algorithm to cricket team selection; however it can be applied to any
multiplayer sports. There are constraints which are specific to sports and this
algorithm is flexible to accommodate those constraints. We have tested the
algorithm on bunch (top) of Pakistani cricket players for test team selection. The
results indicate the usefulness of the proposed algorithm. Extended version of above
mentioned work with complex and specific domain with a flavor of constraint
satisfaction problem solving technique would be more effective. Furthermore the
individual performance of the player may also be included for team selection
process.