HANOI UNIVERSITY

Faculty of Information Technology

SPECIAL SUBJECT 2
LATEX REPORT
Game Theory Application:
Optimizing film release time at movie theatres
for film studios using the NSGA-II algorithm

Instructor: Trinh Bao Ngoc

Students:
Vu Thi Thom - 2101040176
Do Hoang Minh Duc - 2001040034
Trinh To Uyen - 2001040221
Dang Thi Khanh Chi - 2101040037
Ly Thuy Dung - 2101040064
Trieu Truong Giang - 2101040076

Class: Thurday - Group 03

 Game Theory Application: Optimizing film release time
at movie theatres for film studios using the NSGA-II algorithm

Abstract
Determining the optimal film release time is a challenge for film studios because of fierce competition
among stakeholders and seasonal changes in audience moviegoing habits. Film studios may not maximize
box office revenue without optimal release schedules, which can harm the movie’s reputation and future
projects. To surmount these obstacles, Game theory, a mathematical approach analyzing how players
formulate strategies influencing each other’s outcomes, is applied to promote mutually beneficial collabo-
rations between film studios for revenue optimization. By applying Game theory, film studios may select
the ideal release window based on anticipated movements from rivals and demand from viewers. With this
strategic foresight, this study aims to support film studios to face minimal competition, maximize box office
revenue, and foster a thriving film industry. Our approach incorporates Game theory and the NSGA-II
algorithm, a multi-objective optimization technique to find an optimal solution, Nash equilibrium, where
all film studios maximize their profits and have no reason to reverse their course of action. In this case,
film studios can pick the option that best fits their firm interests by using the Nash equilibrium as a trade-
off between maximizing theater and film studio income.
Keywords: film industry, optimize release time, Game theory, NSGA-II algorithm

1 Introduction
Optimizing movie release time is a crucial part in the film industry, allowing for studios to maximize revenue
and reach out to audiences [1]. Studying film launch techniques, [1] emphasized the importance of choosing
the right release time to maximize profit by taking in seasonal demand and industry competition. The impact
of these factors includes the scenario where releasing a film at a time with strong competition can significantly
reduce its chances of success [2]. A prime example is Paramount’s "Star Trek Beyond” (2016) faced major
challenges due to competition from Universal’s "The Secret Life of Pets" and Sony’s "Ghostbusters" [3]. As a
result, "Star Trek Beyond" only grossed $343.5 million worldwide, falling short of expectations. Additionally,
the impact of seasonal demand on film release timing during major holidays such as Christmas, Lunar New
Year often represents a golden opportunity, as theater attendance surges. Based on the data chart [4] in
Figure 1 combined with [5] study, it seemed that films released during holiday periods usually see opening
weekend revenues 70% higher. Therefore, the impact of seasonal demand is extremely important to the film
release timing.

Figure 1: The percentage of box office takings

The release time of a movie significantly influences its box office success and studio reputation, but it also
poses risks, potentially leading to financial losses. As [6] points out, factors like audience preferences, and

1
studio competition all play critical roles. "Ghost in the Shell"’s 2017 launch was a significant setback, as the
US market experienced a significant drop. The film, with a mere 42% rating on Rotten Tomatoes [7], failed
to resonate with US cultural sensibilities and preferences. The incident shows the unpredictable nature of
audience preferences, making accurate forecasting a constant challenge for studios. The industry is gaining
traction due to the rise of new studios and online streaming platforms. Furthermore, traditional studios now
face fierce competition from online streaming services such as Netflix, Disney+, etc., as audiences embrace
these new options [8]. The movie industry is undergoing significant changes, necessitating a robust approach
to information security to maintain audience interest. Improperly managed hazards can lead to significant
financial losses and damage a studio’s reputation.
Film studios face challenges in optimizing film release timing, balancing competition as popular stars,
and extensive promotion tend to attract more viewers [1]. Additionally, studies by [9] show the influence of
the distribution of cinemas on the potential revenue for a film, especially for blockbuster releases. The high
production and marketing expenses necessitate strategic timing of film releases to generate profits. Releasing
during periods of strong box office sales accelerates revenue growth [10]. Audience preferences significantly
influence a film’s success, necessitating market research for studios to cater to these preferences. Figure 2,
derived from a survey of the Indian market by [11], provides information on audience needs and decision-
making processes when choosing movies. Studios face a significant challenge in balancing conflicts to optimize
release timing for increased viewership, revenue, and financial success.

Figure 2: Factors considered while deciding on a movie

In our paper, we’ve opted to employ Game theory and Nash equilibrium to tackle the intricate problem
of optimizing film release time. Game theory is a branch of mathematics and economics, which provides a
framework for understanding how individuals or entities make decisions when their choices depend on the
choices of others [12]. How Game theory works is further explained in the example below Table 1. In this
example, two studios face the decision of choosing suitable time slots for their films. Regardless of what the
other decides, each studio gets better gain when they compete, therefore mutual competition is the Nash
equilibrium solution of this example. Nash equilibrium is a central concept in Game theory[14], referring to
a situation where each player’s strategy is optimal given the strategies of the other players [15]. By finding a
Nash equilibrium solution for the release time problem, we hope to find an optimal solution where every entity
involved in the problem can maximize their benefit. Overall, Game theory and the Nash equilibrium offer
unique and analytical insights for understanding strategic interactions and optimizing solutions in complex
scenarios. Therefore, it’s the most suitable method for solving the problem of optimizing film release time.
Aside from Game theory and Nash equilibrium, in this study, we also utilize a variant of MOEA, the
NSGA-II algorithm, to find the optimal solution for choosing film release time. MOEA stands for Multi-
Objective Evolutionary Algorithm, a type of optimization algorithm used to solve problems with multiple
conflicting objectives [16]. Among the MOEA framework, NSGA-II algorithm stands out because it fol-
lows the basic principles of evolutionary algorithms but incorporates specific mechanisms tailored for multi-
objective optimization [13]. The film release time problem is an intricate issue because it involves balancing

2
 Studio 1
Cooperate (balance the time slots) Compete (get more time slots)
Studio 2
Cooperate (balance the time slots) 1000, 1000 1500, 500
Compete (get more time slots) 500, 1500 500, 500

Table 1: Two-player strategy game with two-by-two matrix

numerous objectives to make strategic decisions that can significantly impact the success or failure of a film
[16]. By incorporating NSGA-II, we can potentially find strategies to maximize box office revenue for film
studios. Therefore, by applying Game theory and the NSGA-II, we hope to effectively handle the conflicting
goals of entities within the film industry, ultimately finding the optimal solution to maximize benefits for all
parties involved.
In order to solve the film release time optimization problem by using a combination of the Game theory
framework and the NSGA-II algorithm, this study raises the following questions:
1. What criteria need to be considered when studios choose film release time?
2. How can we analyze the decision-making process of studios when choosing film release time?
3. How to find the optimal solution to the film release time problem for all film studios?
To make an informed decision regarding the optimal timing for film releases, several crucial factors must
be taken into account. Through observations and extensive literature review, we can categorize factors that
should be considered in the game inputs into the following: level of competition, as films of similar genres
released simultaneously compete not only in terms of quality but also in audience reception and market
share; film’s popularity plays a key role in determining its release schedule. Studios may leverage a trending
genre to ride a wave of audience interest or release a sequel to a previously popular film to maximize box
office revenue. The final criterion is the demand for time slots, films screened during weekday evenings and
weekends tend to generate significant profits.
In the context of selecting the optimal film release timing, we opt for the strategy to analyze and calculate
the decision-making process. Beginning with the Game theory framework, we define studios as players. Based
on their release time choice and the demand of the audience for the time slot, we created strategies for that
film to maximize performance. Through the application of these strategies in diverse real-world scenarios,
we gather data that enables analytical exploration aimed at identifying the Nash equilibrium, a state where
conflicting interests among studios are eliminated.
After defining the competitive landscape and constructing a payoff matrix reflecting audience demand
for different film release windows, our approach utilizes NSGA-II. We handle multiple objectives for studios,
such as maximizing box office returns, minimizing competition, and even fostering a healthy industry balance.
NSGA-II then finds release schedules for each studio, considering the payoff matrix and the impact on all
objectives. Subsequently, Pareto optimum solutions are defined to provide a range of suitable release options.
Ultimately, given the competitive landscape, studios may decide which solution best matches their priorities.
2 Related work
Recognizing the urgency of the problem of optimizing movie release timing in theaters, numerous previous
studies have focused on applying Game theory to optimize film release schedules. However, this problem
remains largely unresolved. A number of related publications have researched this issue. Research articles
have discovered breakthroughs and effective tactics to maximize revenue and reduce studio rivalry by utilizing
methodologies and models. These articles provide us with an answer to the dilemma of when to optimize the
release time.
Choosing a movie release time has been formalized by [1] as an attractive competitive game in which
players (film studios) weigh both prospective revenues and rivals’ decisions. Experiments show that most
studios, accounting for 84% of the market, do not change their release time, suggesting they follow strategies
to reach a Nash equilibrium. [17] used a Gradient Tree Boosting model and a Set Partitioning Problem.
The method combines session demand forecast and schedule optimization, using an iterative procedure to
capture their mutual interdependence. [18] article suggests a U-shaped association between movie popular-
ity and high-demand week release. The results were then demonstrated to be consistent with data from

3
global releases, with the regression coefficients indicating opposite signs for blockbuster and non-blockbuster
cases. [20] presented a 15-year-simulated agent-based market model in 2022, exhibiting comparable worldwide
behavior when movie makers plan films using risk-sensitive techniques.
In their research, [21] analyzed optimal release times based on platform characteristics, movie success
factors, and market factors. Results show that optimal release time increases with ad revenue, broadband
penetration, and piracy rate, while optimal fee reduces non-linearly with release time and depends on OTT’s
risk profile. [22] present a strategy for projecting movie box office using social media data. It gathers online
evaluations from microblogs and derives metrics such as sentiment, customer RFM value, and attentive-
ness. The challenges of scheduling cinema screenings and the limitations of a cinema are discussed by [23].
Regression models were created using statistical data analysis to find predictors of movie attendance.
[24] employed individual suggestions and user-based forecast models in their study, as well as user rating
data, to improve accuracy and performance. [25] also use machine learning approaches in their study work.
They use historical data from IMDb, Rotten Tomatoes, Box Office Mojo, and Metacritic to estimate movie
success rates. [26] investigated the influence of e-WOM on movie ticket sales in 2022, using data from Box
Office Mojo and Rotten Tomatoes as their sources. The results demonstrate a direct association between
ticket sales and internet searches, and online ratings from experts and people also have an impact on sales.
[27] found that internet streaming platforms had as much effect on film production and distribution deci-
sions as traditional studios. Studio executives employ commitment institutional logic to prioritize theatrical
releases and box office receipts, but internet streaming providers use convenience institutional logic to sell
subscriptions.
[28] offer a unique cinema attendance model called DISBM that takes seasonality into account. It in-
tegrates intertemporal demand shift patterns with weekends and vacations, using the Bass model as its
foundation. [29] evaluated short-term cinema attendance forecasting models based on data from 179,103
individual shows in Poland. The results reveal that models based on cinema- and region-specific characteris-
tics, movie attributes, and title popularity are the most successful in forecasting individual show attendance.
[30] introduce a Cinema Ensemble Model (CEM), a unique way to forecast box-office film performance that
outperforms prior models by including a new feature from transmedia narrative theory and employing an
ensemble approach. [31] investigated a sparse rivalry between two studio producers, examining their posi-
tioning, advertising, and quality investments. To investigate genuine market scenarios, it employs both an
analytical and an agent-based modeling approach.
Based on prior research in this field, there have been encouraging advancements in box office revenues
for film studios. However, there are still unresolved challenges in the existing research. Primarily, studies on
the impact of competitive or collaborative relationships between film studios on film distribution are lacking.
Furthermore, while certain studies have addressed factors influencing film release time, no specific methods
or models have focused on this domain. Therefore, the overarching issue lies in the lack of exploration into
optimizing film release timing in theaters. Previous studies primarily focused on objective film factors to
assess box office outcomes upon release, such as genre classification, cast composition, narrative content,
market influences, social media and online platform engagement, among others, to predict audience interest.
Consequently, there is a need for more concentrated and in-depth research to develop methods and models
for optimizing film release timing. This is essential to ensure box office performance and maximize profits in
the competitive and diverse landscape of today’s film industry.
In the ever-changing landscape of the market and fierce competition among online services, determining
the timing of film releases can be a strategy that significantly influences the success of a movie. In order
to improve the issue, this research creatively integrates the NSGA-II algorithm with concepts from game
theory, namely Nash Equilibrium. It provides a methodical way to account for industry changes by filling
in the gaps left by current approaches, increasing profitability by optimizing box office income while taking
audience preferences and rivalry into account. This integration increases the practical use of game theory
in real-world circumstances and revolutionizes film distribution tactics. This study aims to change the way
studios approach release scheduling by offering a strict structure for decision-making that will produce more
profitable and effective results. In an increasingly competitive film business scene, the strategic application
of Nash Equilibrium, bolstered by the computing power of NSGA-II, guarantees optimal outcomes and well-
informed decision-making.

4
 Problem Method Year
Analyze film release time Game theory model 2021
Session demand forecasting, and movie schedule optimization GTB model, ALNS heuristic 2023
Analyze movie release strategy Analysis of regression 2020
Scheduling strategies of studios on the US movie market Agent-based simulation model 2022
Identifying the ideal release window for maximizing profits An analytical model 2019
Forecasting the box offices of movies coming soon based on A method based on improved 2022
social media analysis Bass models
Optimizing the rental schedule Regression models 2023
Predict box office success Machine learning 2017
Predicting movie ticket sales using Google trends In-depth and comprehensive big 2022
data analysis
Analyzing different movie viewing trends based on different Bass model 2017
seasonal effects
Analyzing the interactions between online streaming services Game-theoretic approach 2020
and how they distribute films
Forecasting, and movie schedule optimization Consumer-centric Model 2018
Forecasting Cinema Attendance Linear regression models 2020
Predicting movie success Machine learning 2018
Analyzing the competitive landscape to make movie release Agent-based model (ABM) 2017
decisions

Table 2: Summary of relevant literature according to the applied methods and approaches
3 Problem Definition
Determining the optimal film release time may encounter difficulties, resulting in decreasing box office revenue
and harming the movie’s reputation. Factors that affect the film release time include seasonal changes, fierce
competition, and fixed effects across studios and channels [31]. Therefore, this research focuses on finding the
optimal film release time as an initial step toward overcoming that circumstance. The central technical issue
is to construct and solve a comprehensive game-theoretic model. In this model, film studios as players are
divided into limited sets P = {Pi , .., PM }, where Pi is a specific film studio, and M ∈ N ∗ is the number of
film studios participating in the game. Then, we designate P0 to represent the special player as the audience
and Bi to be the set of characteristics that movie studios can use to assess the release timing of a film.
Moreover, strategic decisions are encapsulated by strategy, denoting the chosen actions of each stakeholder.
The objective of this game-theoretic model is to optimize these strategies to maximize the respective utilities
of the stakeholders, considering the interplay between characteristics. The Nash equilibrium solutions of the
game model signify stable points where no studio gains an advantage by changing its release date alone [32].
These solutions offer valuable insights into striking a balance when selecting a film release date that aligns
with the interests of all parties.
To illustrate the complexities of film release time decision-making, consider a hypothetical scenario based
on recent events. Four major studios, Universal Pictures (UP), Warner Bros. (WB), Disney (DI), and
Paramount Pictures (PP), as rational players, are preparing for summer release. Each film studio aims to
maximize its revenue; however, head-on competition within the same genre during peak seasons can lead to
lower studio revenue for all parties. Table 3 captures a hypothetical dataset of variables for this scenario.
Our approach addresses the complexities of determining film release time by using a game-theoretic
model that analyzes strategic decision-making by film studios considering competition and various market
factors. Key players such as Universal Pictures, Warner Bros., Disney, and Paramount Pictures are identified,
providing necessary input for the subsequent Game Modeling Module. In this instance, a comprehensive

5
 Variable UP WB DI PP
Film The Spy Next Wonder Twins Enchanted Paranormal
Door Kingdom Activity 5
Film portfolio 1 1 1 1
Past performance 1 1 1 0
Production cost 120 150 130 80
(million USD)
Popularity degree 2 1 3 4
Competitive level 3 2 1 4
Estimated revenue 180 220 250 110
Demand for time 4 5 4 2
slots
Strategic Maximize studio Collaboration Fair film release Adaptive
Preference revenue time strategies

Table 3: Hypothetical Dataset of Key Variables

game matrix is constructed, encompassing the strategies and payoffs for every participant, impacted by
factors such as film j, past performance H, production cost P , popularity degree D, competitive level C,
estimated revenue R, demand for time slots T , and potential profit P r. Then, the model incorporates
Nash equilibrium criteria to guarantee stability in the strategic interactions. MOEAs optimize equilibrium
solutions, considering objectives related to stability and mutual benefits, iteratively exploring the solution
space, evaluating different combinations of strategies, and evolving towards configurations that represent
optimal Nash equilibria. This process contributes to the precision and efficiency of the overall approach,
ensuring that the identified equilibrium points align with the complex considerations inherent in determining
film release times. After that, we continue to analyze and evaluate the outcomes, acting as a crucible where
theoretical frameworks of Game theory and optimization algorithms converge with the results obtained.
This review provides a comprehensive understanding of how film studios determine film release times in
the current landscape of the film industry. This approach, visualized through the flow diagram, provides a
structured and rigorous framework for navigating the complexities of deploying advanced satellite technology
in conflict-prone regions.

Figure 3: Flowchart of approaching the problem of determining film release time

Within the framework of construction project scheduling and achieving Nash equilibrium, the key players
are film studios. The characteristics of these players are multifaceted:
• Film (j): the specific film that the studio will release.
• Film portfolio (F Pi ): the total number of films the studio will release.
• Past performance (Hi ): the historical data and insights on the studio’s endeavors, denoted by Hi ∈ [0, 1],
where Hi = 1 for a studio with the highest profitable rate, and Hi = 0 for a studio that is least likely
to make profit.
• Production cost (Pj ): the production cost of the film j.
• Popularity degree (Dj ): Reflecting the level of public interest and engagement with the film, denoted
by Dj ∈ [0, 1], where Hi = 5 for a very popular film, and Dj = 1 for an unpopular one.
• Competitive level (Cj ): the competitive level of the film, the higher the value the more likely the
audience will choose to watch the film.
• Estimated revenue (Rj ): The estimated revenue of the film j.

6
 • Demand for time slots (Tj ): the audience demand of time slots, denoted by Tj ∈ [1, 5], where Tj = 5
for a time slot with the highest audience demand, and Tj = 1 for a time slot with the lowest audience
demand.
• Potential profit (P rj ): the potential profit of the film j.

Players Characteristics Strategies

Maximizing competitive level of the film Cj by aiming for
Film studios Film (j)
the most in demand time slot (Str1 )
Past performance (Hi) Cooperating about the in demand time slots Tj (Str2 )
Production cost (Pj)
Popularity degree (Dj)
Competitive level (Cj)
Estimated revenue (Rj)
Demand for time slots
(Tj)
Potential profit (Prj)

Table 4: The table summarizes the characteristics of the players

In this strategic context, we apply Nash equilibrium to help studios make strategic film release decisions
while considering the competitive landscape. This requires systematic and sequential integration of the Nash
equilibrium with the NSGA-II algorithm to give the most optimal results for the studios. Initially, we collect
the necessary inputs to create a set of players. Next, we consider various variables affecting the film’s box
office success and the studio’s potential revenue. The following characteristics: j, H, P , D, C, R, T , and
P r are supposed to determine specific strategies for each player involved in the film release time game. This
decision is based on the actions and strategies of other competing studios in the game model and aims to
determine the optimal time for the player. Once the payoff function and strategy have been determined, the
next step is integrating the NSGA-II algorithm and Nash equilibrium. After the NSGA-II algorithm generates
a set of Pareto optimal solutions, Nash equilibrium will come into play to further refine the decision-making
process to optimize film distribution. The objective is to find stable release dates for each player where no
studio is incentivized to change its chosen date considering other studios’ strategies.

Figure 4: Flowchart for Game theory-based optimized film release time solution
The diagram outlines a method for optimizing movie studios’ release times. It involves identifying movie
attributes, collecting data, setting parameters for the NSGA-II algorithm, creating random release schedules,
evaluating suitability based on revenue, competition, and audience demand, applying genetic operators, and
analyzing results.

7
4 Models
In the research applying Game theory to IT project management, Dr. Trinh Bao Ngoc experimented with
and proposed the Unified Game-based Model, a comprehensive game theory model for various conflict classes,
based on a common structure for imperfect information games, cooperative games, and non-zero-sum games
[33]. It provides an adaptable framework for analyzing and modeling challenging decision situations in
dynamic multi-agent environments. By integrating Game theory with optimization algorithms like NSGA-II,
it offers notable advantages, particularly in improving film release scheduling at movie theaters.
Using a Unified Game-based model approach (Trinh Bao Ngoc, 2020), integrating Game theory and
optimization algorithms, we offer an organized framework for examining decisions about the release schedule
of films as a strategic game between studios to produce fair and effective results.
The Unified Game-based model proposed to solve this problem is:
G = <{P0 , P }, {S0 , S}, {u0 , ui }, Rc >
In which:
• G: represents our game model

• P0 represents the special player as the audience, and is a practical factor because the audience’s con-
sumption habits, preferences, needs and feedback can impact the release time of films.
• S0 = {s01 , .., s0j , .., s0M 0 } is the set of strategies of the audience P0 , where M0 is the number of
the audience’s strategies; s0j is the audience’s information structure such as: consumption habits,
preferences, needs and feedback.
• P = {p1 , .., pi , .., pN } represents the film studios, where N is the number of participating studios.
• Si = {si1 , .., sij , .., siM i } is the set of strategies of player i(1 ≤ i ≤ N ), possessing characteristics. Here,
s0i is the ith strategy of player P , and Mi is the total number of available strategies for player i.
• u0 : s0i → ℜ is the payoff function of special player P0 , refers to player’s strategies S0 to real numbers
ℜ.
• ui : S1 → ℜ is the payoff function of player i, referencing the player’s strategy to a real value.
• Rc is the vector space representing the set C of conflicts present in the film release problem. Here, a
vector ⃗v ∈ Rc represents a conflict between M players (1 ≤ M ≤ N ), occurring among the strategies of
each film studio, and can be described as (spq , .., sxy ), with spq , sxy ∈ Si , (1 ≤ p, x ≤ N ), (1 ≤ q ≤ Mp )
and (1 ≤ y ≤ Mx ).
To construct a payoff function for the film studios, first we present a mathematical equation for the weighted
score of the competitive level of the film:
D
Cj = Tj × P j Dj (1)
k∈N
Where:
• Cj : the intensity of competition the film faced within the market.

• Dj : the level of public interest and engagement with the film j.

• Tj : the audience demand of time slots.
• N : the number of films that choose the same time slots.
Then, we have the equation to calculate the potential profit of each film:
P rj = (Hi × Dj × Cj × Rj ) − Pj (2)
Where:
• P rj : the potential profit of the film j.
• Hi : the historical data and insights on their previous film endeavors.

8
 • Rj : the estimated revenue of the film j.
• Pj : the production cost of the film Fi
Following that, we have the payoff function for each
PK filmPstudio i:
rj
ui = j=1 Rj −P j
(3)
Where:
• ui : the payoff function of each studio.
• K : the total number of films released from each studio.
Then we have payoff function for the special player - the film audience:
PM PK Hi ×Dj )
(
(4)
i j Tj
uo = M
Where:
• M : the total number of film studios
• K: the total number of films released from each studio.
The equations above are essential components within the Unified Game-Based model to distribute time
slots at movie theaters systematically to achieve profitable revenue for film studios. These four equations
combine all the significant characteristics of each film studio in the procedure, thereby establishing a fair and
efficient system. Equation (1) determines the competitive level of the films. This equation includes the two
key characteristics T and D. Equation (2) finds the potential profit of each film using the characteristics H,
D, C, R, and P . By calculating the total potential profit of films and comparing it to the estimated profit
of each studio, we can determine the payoff of each studio using equation (3). If ui < j, it means the studio
has suffered a loss in profit; and if ui ≥ j, it means that the studio has made as much as or even more than
expected profit. Equation (4) is the payoff function for the special player - the audience. The audience wants
to easily buy tickets for popular films from well-known studios, with not too much competition, therefore the
equation includes the characteristics H, D, C and T .
Utilizing a clearly defined and robust computational approach of the Unified Game-based model has several
significant benefits. Firstly, the unique approach of using the Unified Game model offers a new understanding
of the interactions and conflicts among stakeholders within the film release game. Secondly, our models
provided a numerical evaluation of each characteristic, enabling data-driven comparison. Thirdly, compared
to other research about the same topics, our model has considered a broader range of characteristics regarding
films and film studios. Furthermore, by using the NSGA-II algorithm, our model can navigate complex
decisions and identify Pareto-optimal solutions, enhancing the computational efficiency of film release time
optimization. Lastly, the use of payoff functions in our model provides a quantitative basis for analyzing and
comparing strategies, simplifying the decision-making process and enabling for more data-driven insights.

Figure 5: Flowchart of the Unified Game-based model

The process of optimizing film release times can be modeled as a non-cooperative game, with the process
of solving the model as seen in figure 4. The first step involves developing the game structure and the
payoff function for each studio. Then, a Nash equilibrium is identified, a stable state where no studio has
an incentive to change their decisions. This signifies the optimal release time for each film studio, while
considering the payoff for the audience. NSGA-II emerges as a crucial tool to solve this efficiently. The
algorithm aims to find out the solution that strikes a delicate balance between the goals of the entities,
guaranteeing profitable revenue between film studios. By carefully weighing the trade-offs, NSGA-II provides
a pathway to optimization, ensuring both box office success and advancement in the film industry.
9
5 Algorithm applied for the problem
An NP-hard (Non-deterministic Polynomial-time hard) problem is a type of computational problem for
which no known efficient algorithm exists to solve it in polynomial time. Optimizing film release time
(OFRT) is considered an NP-hard problem because it is solved in polynomial time OOF RT = xn and n is a
constant (n ∈ N + ). The decision of when to release a film involves balancing numerous factors, including
maximizing box office revenue Rj , minimizing competition with other releases Cj , and aligning with audience
preferences Dj and market trends. When the size N of participating studios P = p1 , .., pN and Mi players
of Si = si1 , .., siM i escalates, it becomes more difficult to optimize the solutions. However, that issue can
be solved well by the NSGA-II algorithm. According to [34], the time complexity at which NSGA-II runs is
O(M × N 2 ), where M is the number of objectives, and N is the population size.
The NSGA-II algorithm is an effective tool for matching film studios P = p1 , .., pN with appropriate
films Si = si1 , .., siM i . NSGA-II is a great multi-objective optimization technique for testing many superior
solutions in difficult search areas. In this study, priority lists for both film studios pi and films sj were
created by applying NSGA-II. It successfully navigates a complicated environment of competing goals, such
as revenue maximization (Rj ), competition minimization Cj , and audience preferences Dj . The trade-offs
between these potentially opposed goals are successfully investigated using NSGA-II, which frames the issue
as a multi-objective optimization assignment. It can provide a set of Pareto-optimal solutions, which best
balance the various needs of films and film studios. The flexibility of NSGA-II to manage multiple objectives
makes it appropriate for the film business, which is dynamic and unpredictable.
NSGA-II seeks to define the chromosome and fitness function to solve the Genetic Algorithm’s multi-
objective optimization issue. In optimizing the film release time problem, we can identify the strategies for
four film studios UP, WB, DI, and PP, then represent them as chromosomes. The figure below is an example:

Where:
• Each chromosome encodes a string of strategies and appears as a sequence of genes, wherein each gene
denotes the specific strategy chosen by the player.

• The number of genes is equal to the number of all strategies for each player.
• The value 0 signifies the absence of a strategy choice, while values 1, and 2, denote specific types of
strategies selected by the players.
The process flow of the chromosomes to effectively encapsulate and transmit genetic information that
represents the best solution is shown in Figure 6. During crossover, segments of chromosomes from parent
individuals amalgamate to engender new offspring chromosomes. Meanwhile, mutation entails stochastic
alterations to one or more genes within a chromosome. Following each evolutionary iteration, individuals
undergo evaluation contingent upon the efficacy of their strategies in the games played. Those exhibiting
superior performance are retained for subsequent generations, while less proficient individuals may go extinct.
Besides, the amalgamation of chromosomes with fitness functions serves to unveil optimal or adaptive
strategies within a specified game milieu. Each chromosome symbolizes an individual’s strategy within the
population, while the fitness function gauges their performance based on predefined criteria, encompassing
metrics like point accrual or win rates. Throughout each evolutionary cycle, the fitness function is employed
to scrutinize every individual within the populace. Individuals boasting elevated fitness values are accorded
precedence and stand a heightened likelihood of being chosen as parental candidates for the subsequent gen-
eration. Through crossover and mutation operations, novel offspring are engendered, inheriting advantageous
traits from their progenitors. With each ensuing generation, the population tends to harbor individuals har-
boring superior or more adaptive strategies, thereby engendering optimization or adaptation to the game
environment.

10
 Figure 6: The process of the chromosome in evolutionary game theory

The fitness function is formally defined as follows:

PN
u +u
uall = i=1N i 0 (5)
Where:

• u0 is the payoff function for special player

• ui the payoff function of each film studio
• N is the number of films to be released in a given period
The algorithm tackles a problem by defining input values as chromosomes and using the Tournament
Selection technique to compare random chromosomes. Crossover and mutation occur to create new solutions,
with the degree of mutation changing based on newborn conditions. The inferior solutions are eliminated to
maintain the total solutions. The process continues until the set of solutions remains stable after iteration or
a condition is met.

Figure 7: The process flow of the algorithm

NSGA-II can be applied to solve the determining the film release time as a multi-objective optimization
problem. In the context of film distribution, the main objectives could include minimizing competition or
maximizing box office revenue. The algorithm helps find a set of solutions that represent a trade-off among

11
these conflicting objectives, known as the Pareto front.

Input: P is the population - set of all possible solutions

N is the size of the population
Output: S is the survivors - set of solutions (chromosomes) that pass the condition
begin
1 Initialize population P of size N
2 The value of the initial population F(P) = EvaluateValue(P)
3 Fronts = FastNonDominatedSort(P, F(P))
4 For each front F in Fronts:
5 parent_generation newP = TournamentSelection(F)
6 survivors S = newP
7 while (isFluctuated(F(S)) = True):
8 crossover_child C = CrossOver(newP)
9 mutation_child M = Mutation(O, C)
10 S.append(M)
11 Evaluate the value of survivors F(S)
12 if isFluctuated(F(S)) = False:
13 break
14 end for loop
15 return set of survivors S in Fronts
end.

6 Computational analysis
This section outlines the system configuration used in our experimental setup and presents results that
contribute to evaluating the effectiveness of the game-theoretic algorithm used in the paper. The testing
environment includes Windows 11, Intel Core i5-11320H, 2.50 GHz, and 8GB of RAM. In addition, we
specifically add algorithm parameters in the following table:

Experiment Population Number of Maximum Selection Crossover Mutation

Parameters size Generations Iterations Rate Rate Rate
Value 100 100 100 0.8 0.7 0.01

Table 5: The NSGA-II experimental parameters for film release time optimization
In this context, we provide a dataset containing weighted values for different factors affecting film release
timing optimization. We provide a dataset that includes information from over 100 films produced by 50
distinct film studios, labeled alphabetically. Each film focuses on two strategies and is characterized by
various performance metrics, including past performance, related costs, and stakeholder demand. Table 6
below is a partial representation of the dataset relative to the studios:

Characteristics
Studios Strategies Film j
Pj Rj Prj
Hi Dj Tj Cj
(M) (M) (M)
1 4 5 0.32 120 450 296.48
Str1
2 3.5 4 0.28 150 310 73.26
A 0.73
1 4 2.5 0.26 120 450 219.64
Str2
2 3.5 3.5 0.27 150 310 65.15
3 3.5 4.5 0.29 95 345 134.27
Str1
4 4 5 0.32 120 390 201.40
B 0.65
3 3.5 3 0.26 95 345 110.17
Str2
4 4 2 0.25 120 390 130.24

12
 5 3 5 0.29 90 350 151.68
Str1
6 4 4.5 0.31 75 250 169.21
C 0.80
5 3 2.5 0.24 90 350 114.84
Str2
6 4 1.5 0.24 75 250 113.07
Table 6: A part of the dataset for film studio

After extensive data processing, selection, and division among various players and strategies, we have
achieved the following results that maximize their payoff values in Table 7 :

Studio Strategy Payoff value

A Str1 0.5775
B Str1 0.63448768
C Str1 0.551199839

Table 7: Payoff values for strategies

The fitness value calculation results are based on the previously film information dataset and film at-
tributes dataset, using various algorithms described in Table 8 below.

NSGA II eMOEA VEGA

Iteration Fitness Time Fitness Time Fitness Time
value (sec) value (sec) value (sec)
1 0.20636 1.586 0.20100 0.51 0.20305 0.882
2 0.20457 1.26 0.20794 0.464 0.20977 0.94
3 0.20632 1.12 0.20393 0.481 0.20474 0.889
4 0.20769 1.29 0.20544 0.453 0.20351 0.949
5 0.20697 1.101 0.20526 0.489 0.20170 0.987
6 0.20948 1.341 0.20049 0.448 0.20179 1.054
7 0.20702 1.218 0.20039 0.484 0.20463 0.88
8 0.20838 0.95 0.20787 0.486 0.20196 0.982
9 0.20479 0.919 0.20328 0.472 0.20176 0.958
10 0.20499 0.973 0.20589 0.406 0.20880 0.974

Table 8: Comparison between the fitness values and the runtimes of different algorithms.
Figure 8 below, a dot plot chart, showcases the runtime comparison of different algorithms, including
NSGA-II, MOEA, and VEGA, over ten iterations. These runtime values provide insights into the computa-
tional efficiency and processing speeds of each algorithm in the context of the optimization task.

Figure 8: Comparison between the runtimes of different algorithms on the dataset

13
 To show the difference in the fitness values across algorithms, a line chart with markers was created. It
represents the fitness values achieved by each algorithm (eMOEA, NSGA-II, and VEGA) across iterations
using distinct markers. The x-axis shows the iteration numbers, while the y-axis represents the corresponding
fitness values.

Figure 9: Comparison between the fitness values of different algorithms on the dataset
Important quantitative insights were obtained from the experimental results of comparing three multi-
objective optimization algorithms: NSGA-II, eMOEA, and VEGA, using a dataset of 100 individuals. Ten
simulation runs per algorithm, 100 generations, and standard parameters were utilized. In general, the fitness
values concentrate around the 0.2 value, but with different runtimes with eMOEA as the fastest. Although
eMOEA delivered results at the fastest pace, its fitness values proved unstable, with specific instances where
the fitness values reached a new low with 0.20039, while there are instances where it escalated to 0.20787.
While VEGA algorithm has a better average runtime than NSGA-II, its fitness values mostly are low, reaching
about 0.202, with rare occasions shooting up to 0.208. On the other hand, NSGA-II demonstrated remarkable
stability with fitness values, staying consistently around 0.205 in most of the 10 iterations. These experimental
results underscore the stability and efficiency of NSGA-II as a reliable tool for strategic decision-making in
optimizing film release time.

7 Conclusion
Determining the optimal release time for films has always posed a challenge for film studios and the film
industry as a whole. In this study, we utilized Game theory and NSGA-II to determine the best release
times for films from various studios, aiming to maximize each studio’s profits and minimize competition.
Although our approach offers a novel solution with clear quantitative outcomes, it has limitations as it does
not account for all the unique characteristics of films and film studios, nor does it consider the role of movie
theaters. In our research, we created models to calculate the payoff of each studio, how their payoff affects
the audience, and how good the solution is using the fitness function. All of the models and calculations
allow film studios to find the optimal release time for films, helping studios maximize their potential profits.
From the models and experiment results, we found that the NSGA-II algorithm is the best option because of
its stable calculation, which makes it the perfect algorithm to identify the optimal solution to the film release
time problem. This study introduces a new quantitative approach to choosing film release time, providing a
foundation for future research by academics and film industry professionals seeking to optimize film release
schedules.

References
[1] Liu, A. H., Chen, M., Wang, X., Wang, Z. (2020). A game-theoretical approach to analyze film re-
lease time. 2020 IEEE 32nd International Conference on Tools With Artificial Intelligence (ICTAI).

14
 https://doi.org/10.1109/ictai50040.2020.00135
[2] Fu, J. (2023). Empirical analysis on the influential factors of business strategy in the U.S. movie industry
based on marketing mix models and SWOT analysis of data about market performance A case study on
DC Films. European Union Digital Library. https://doi.org/10.4108/eai.18-11-2022.2326903
[3] Mendelson, S. (2016, July 31). Friday box Office: “Star Trek” plunges 69%, “Ghostbusters” top-
ping $100M. Forbes. https://www.forbes.com/sites/scottmendelson/2016/07/30/friday-box-office-star-
trek-plunges-69-ghostbusters-topping-100m/?sh=43eadd437fce
[4] Limited, A. A. (n.d.). Seasonal changes in the genres of box office re-
leases. https://ampereanalysis.com/insight/seasonal-changes-in-the-genres-of-box-office-
releases?fbclid=IwAR1gDQ5q8apSRyVkXnScB7oKzLta0k_Z5_E1d5tyzuae-MxlHe4_9jbgS60
[5] Hao, B. (2023). The Analysis of the Factors that Influence the Film Revenue. Highlights in Science,
Engineering and Technology, 47, 154–159. https://doi.org/10.54097/hset.v47i.8184
[6] Stimpert, J. L., Laux, J. A., Marino, C., Gleason, G. (2011). Factors influencing motion pic-
ture success: Empirical review and update. Journal of Business & Economics Research, 6(11).
https://doi.org/10.19030/jber.v6i11.2488
[7] Ghost in the Shell | Rotten Tomatoes. (n.d.). Rotten Tomatoes.
https://www.rottentomatoes.com/m/ghost_in_the_shell_2017
[8] Kasper, A. (n.d.). A New Studio Era: How Netflix is Becoming a Threat to the Film Industry.
PDXScholar. https://pdxscholar.library.pdx.edu/honorstheses/1271/
[9] Lotman, E., Voodla, A., & Uusberg, A. (2019). What’s the value of a star? Actor familiarity and
likeability effects on emotional mimicry of cinematic displays. Baltic Screen Media Review./Baltic Screen
Media Review, 7(1), 4–16. https://doi.org/10.2478/bsmr-2019-0001
[10] Einav, L. (2007b). Seasonality in the U.S. motion picture industry. the RAND Journal of Economics/the
Rand Journal of Economics, 38(1), 127–145. https://doi.org/10.1111/j.1756-2171.2007.tb00048.x
[11] Nimbagal, Suvarna. (2023). PHRONIMOS: THE KIAMS JOURNAL Factors leading to the adoption of
OTT services-An analysis of the impact of OTT on traditional media. 3. 30-47.
[12] Rubinstein, A., & Osborne, M. J. (1994). A Course in Game Theory. Massachusetts: MIT Press.
[13] Deb, K. (2001). Multi-objective optimization using evolutionary algorithms. John Wiley & Sons.
[14] Nash, J. (1951). Non-Cooperative Games. Annals of Mathematics, 54(2), 286–295.
https://doi.org/10.2307/1969529
[15] Dixit, A. K., & Nalebuf, B. J. (2008). The art of strategy: A game theorist’s guide to success in business
and life. W. W. Nortan.
[16] Coello, Carlos & Veldhuizen, David & Lamont, Gary. (2007). Evolutionary Algorithms for Solving Multi-
Objective Problems Second Edition. https://doi.org/10.1007/978-0-387-36797-2
[17] Groen, M. (2023). The Movie Scheduling Problem: Forecasting and Optimization. Econometrie.
http://hdl.ha0278 ndle.net/2105/7
[18] Cabral, L., & Natividad, G. (2020). Movie release strategy: Theory and evidence from international
distribution. http://dx.doi.org/10.1111/jems.12344
[19] Bertolotti, F., & Roman, S. (2022). Risk sensitive scheduling strategies of production studios on the US
movie market: An agent-based simulation. http://dx.doi.org/10.3233/IA-210123
[20] Sharma, M., Basu, S., Chakraborty, S., & Bose, I. (2022, 12). Determining the optimal release time of
movies: A study of movie and market characteristics. https://doi.org/10.1016/j.dss.2022.113893

15
[21] Zhang, C., Tian, Y.-X., & Fan, Z.-P. (2022). Forecasting the box offices of movies coming soon using social
media analysis: A method based on improved Bass models. https://doi.org/10.1016/j.eswa.2021.116241
[22] Asnina, N., Fateeva, I., Yakovenko, N., & Stativa, E. (2023). Methods of data analysis in the problem
of optimizing the rental schedule. http://dx.doi.org/10.1051/e3sconf/202338906030

[23] Ruhrländer, R. P., Boissier, M., & Uflacker, M. (2018). Improving Box Office Result Predictions for
Movies Using Consumer-Centric Models. http://dx.doi.org/10.1145/3219819.3219840
[24] Quader, N., Gani, M. O., Chaki, D., & Ali, M. H. (2018). A Machine Learning Approach to Predict
Movie Box-Office Success. http://dx.doi.org/10.1109/ICCITECHN.2017.8281839

[25] Anuja, S., Aditya, Y., & Shiv Kumar, S. (2022). Predicting Movie
Ticket Sales Using Google Trends: Implication of Big Data Analyt-
ics. https://www.proquest.com/openview/4ba7d10430d76f6b3fb4d4438cf80b1c/1?pq-
origsite=gscholar&cbl=54462
[26] Hadida, A. L., Lampel, J., Walls, W. D., & Joshi, A. (2020). Hollywood studio filmmaking in the age of
Netflix: a tale of two institutional logics. https://doi.org/10.1007/s10824-020-09379-z
[27] Zhang, X., Hou, G., & Dong, W. (2017). Modeling movie attendance with seasonality: evidence from
China. http://dx.doi.org/10.1080/13504851.2017.1279260
[28] Baranowski, P., Korczak, K., & Zając, J. (2020). Forecasting Cinema Attendance at the Movie Show
Level: Evidence from Poland. http://dx.doi.org/10.2478/bsrj-2020-0006
[29] Lee, K., Park, J., Kim, I., & Choi, Y. (2018). Predicting movie success with machine learning techniques:
ways to improve accuracy. https://link.springer.com/article/10.1007/s10796-016-9689-z
[30] Delre, S. A., Panico, C., & Wierenga, B. (2017). Competitive strategies in the motion picture industry:
An ABM to study investment decisions. http://dx.doi.org/10.1016/j.ijresmar.2016.05.005

[31] Nelson Granados, J. M. (2018). Popcorn or Snack? Empirical Analysis of Movie Release Windows.
Pepperdine University, Graziadio Working Paper Series. Paper 1
[32] Federico Paudel. (2013). Optimization in Game Theory: Finding Equilibrium Solutions. Global Journal
of Technology and Optimization, Volume 14:5, ISSN: 2229-8711

[33] Trinh Bao Ngoc. (2020). AP DUNG LY THUYET TRO CHOI VA CAN BANG NASH XAY DUNG
PHUONG PHAP MO HINH HOA XUNG ÐOT TRONG QUAN LY DU AN ÐAU TU CONG NGHE
THONG TIN VA THU NGHIEM TRONG MOT SO BAI TOAN ÐIEN HINH
[34] Deb, K. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II.
https://doi.org/10.1109/4235.996017

16