CSV: Visual Support for Understanding Card Synergy in

Digital Collectible Card Games

Yicheng Xue

and Hiroshi Hosobe

Graduate School of Computer and Information Sciences, Hosei University, Tokyo, Japan

Faculty of Computer and Information Sciences, Hosei University, Tokyo, Japan

Keywords:

Graph Visualization, Video Game, Digital Collectible Card Game.

Abstract:

Digital Collectible Card Games (DCCGs) are a popular genre of video games that typically feature a contin-

uously expanding pool of cards, requiring players to construct their own decks in order to play. However, the

complexity of the game rules and the large number of cards cause information overload, resulting in various

issues. In this research, we propose a framework to help users overcome information overload by providing

a clear visualization of card synergies using 3D graphs. We employ text analysis and the co-occurrence net-

work to calculate synergy scores between cards, and then represent the cards as nodes and their synergies as

edges in our integrative 3D graph. In our experiment, we collected the decks of elite players as our dataset

and visualized the synergies among approximately 1,000 cards in “Yu-Gi-Oh! Master Duel”. To evaluate our

framework, we conducted a questionnaire survey and a usability test with people experienced in playing DC-

CGs. The results indicate that our framework effectively assists users in deck construction and understanding

of the game, and also provide valuable insights for the further development into a full-scale supporting tool.

1 INTRODUCTION

Digital collectible card games (DCCGs) are a popu-

lar genre of video games that typically emulate col-

lectible card games (CCGs, sometimes also known as

trading card games) on digital platforms. Because of

the advantages of being digital, the game mechan-

ics of DCCGs are usually more creative and ﬂexi-

ble. Although traditional CCGs remain very popular,

the popularity of DCCGs has increased signiﬁcantly

over the past few years, reaching millions of play-

ers around the world. “Hearthstone” and “Magic: The

Gathering” are two famous DCCGs that have been all

the rage in recent years (Turkay and Adinolf, 2018).

DCCGs inherit the characteristics of traditional

CCGs, which require players to build a personalized

deck to engage in turn-based one-versus-one matches.

In the same way as CCGs, deck building and match

gameplay are two key components that constitute the

core experience of DCCGs (e Silva Vieira et al.,

2024). However, DCCGs also inherit some of the is-

sues of CCGs, such as information overload (decision

making challenges due to too much information) and

power creep (updates causing older cards to become

underpowered) (Zuin et al., 2022).

Since the playability of CCGs largely depends on

the variety of cards, both DCCGs and CCGs typi-

cally have a large pool of cards that are frequently

updated. The extensive libraries of available cards,

often numbering in the thousands, can be overwhelm-

ing and confusing for beginners who are not yet fa-

miliar with the games. A case in point is the well-

known title “Yu-Gi-Oh! Master Duel”, which boasts

a card pool that exceeds 10,000 unique cards (Konami

Digital Entertainment, 2021). Although many card

databases offer powerful search functions, inexperi-

enced players may still struggle to identify potential

card combos due to the overwhelming volume of in-

formation, where a combo refers to a sequence of ac-

tions that provides signiﬁcant beneﬁt. From the per-

spective of game designers, the issue of imbalanced

card designs often arises due to power creep, leading

to a continuous cycle of needing stronger cards to re-

main competitive (Zuin et al., 2022). Even though

game designers utilize mathematical models to main-

tain the balance of card values, the additional effects

resulting from interactions between cards are some-

times overlooked, which can lead to certain cards be-

coming overpowered. To address this issue, designers

often implement a ban list or make direct modiﬁca-

tions to card effects. However, there is still room for

improvement in the evaluation process during the de-

Xue, Y. and Hosobe, H.

CSV: Visual Support for Understanding Card Synergy in Digital Collectible Card Games.

DOI: 10.5220/0013256900003912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2025) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 611-618

ISBN: 978-989-758-728-3; ISSN: 2184-4321

611

sign phase (Zuin et al., 2022; Vieira et al., 2020).

The main approach to addressing these issues typ-

ically involves the use of artiﬁcial intelligence (AI)

based on simulated combat data. However, these AI

models are usually black boxes, which means that

users do not signiﬁcantly deepen their understanding

of the cards or the game. Moreover, due to the com-

plexity of DCCGs, the performance of these AIs still

has considerable room for improvement (Zuin et al.,

2022; Vieira et al., 2020). Since information visual-

ization presents complex data in an intuitive format,

which could reduce information overload and enable

quicker understanding and decision-making, our goal

is to provide visual support based on transparent an-

alytical methods to enhance users’ comprehension of

the game (Strother et al., 2012).

In this research, we introduce a novel framework

designed to visually support users in understanding

card synergies within DCCGs. A card synergy refers

to the strategic interactions between cards that pro-

duce a combined effect greater than their individual

abilities, a core element of gameplay and deck build-

ing (Dockhorn and Mostaghim, 2019). For our re-

search methods, we use the deck data of elite play-

ers as the dataset to calculate synergy scores be-

tween cards based on card effect analysis and the co-

occurrence network technique. We then conceptual-

ize cards as nodes and their quantiﬁed synergies as

edges to construct an interactive 3D graph visualiza-

tion. In the end, our results of subjective evaluations

indicate that our visualizations effectively help users

understand the synergy between cards, which demon-

strates the feasibility of this research.

2 RELATED WORK

2.1 Research on CCGs and DCCGs

There is limited research on card synergies and re-

lationships in CCGs and DCCGs. Most studies ad-

dressing challenges in CCGs and DCCGs focus on

training powerful AI. For example, a framework uses

deep learning to recommend resource scaling for bet-

ter game balance (Zuin et al., 2022). It combines

neural networks with gradient-boosted decision trees

to predict card values and employs explanation tools

to help developers understand the factors inﬂuenc-

ing these predictions. In the context of AI agents

for CCGs, deep reinforcement learning has been ap-

plied to optimize the drafting process, while neural

networks are used to develop draft agents capable of

building competitive decks (Vieira et al., 2020). Ad-

ditionally, “Q-DeckRec” is a fast deck recommenda-

tion system for CCGs that applies Q-learning to sug-

gest optimal decks based on player preferences and

game state, which signiﬁcantly enhances the deck-

building experience with quick and personalized rec-

ommendations (Chen et al., 2018). These studies are

part of mainstream research in CCGs, focusing on AI

applications to assist users, with promising results.

However, they provide limited support for our visual-

ization research because of differing perspectives on

problem solving.

2.2 Co-Occurrence Network

Since words in context and cards in DCCGs share

similar properties, we can conceptualize cards and

decks in DCCGs as analogous to words and sen-

tences in linguistic structures. To analyze the relation-

ships among these entities, natural language process-

ing techniques may offer valuable insights for our re-

search. The co-occurrence network is a technique for

text analysis that calculates the co-occurrence of enti-

ties and often utilizes graphic visualization to uncover

potential relationships among entities within written

content. In terms of applications, co-occurrence net-

works have been used to analyze Twitter data from

a sample of 3,000 tweets (Puerta et al., 2020). The

analysis indicates that co-occurrence networks, es-

pecially those from pre-processed text, reveal struc-

tural relevance among terms, offering valuable in-

sights for broader text analysis. Another study ex-

amined political tweets related to Hillary Clinton’s

2016 presidential campaign using co-occurrence net-

works to explore the sentiment and structural proper-

ties of word relationships (Fudolig et al., 2022). By

constructing networks with nodes as words and edges

as co-occurrences, the study illuminates connections

between words and sentiments, and reveals complex

patterns and clusters within the data, which demon-

strates the utility of co-occurrence networks in identi-

fying hidden word relationships. Given its effective-

ness in revealing relationships between entities, we

incorporate co-occurrence network analysis as part of

our methodology to construct a relationship map for

cards in DCCGs.

2.3 3D Force-Directed Graph

Since DCCGs have numerous cards and complex re-

lationships, we seek a visualization method that can

handle large amounts of information. 3D graphs, with

their strong interactivity and intuitive presentation, of-

fer an ideal solution. Among the various types of 3D

graphs, 3D force-directed graphs stand out for their

excellent performance in visualization research. An

HUCAPP 2025 - 9th International Conference on Human Computer Interaction Theory and Applications

612

example of 3D visualization in action is a web-based

tool designed to explore scientiﬁc literature in 3D

space (Swacha, 2021). It represents papers as nodes

and citations or co-occurrences as edges, enabling

researchers to analyze complex networks. This ap-

proach enhances the ability to detect relationships and

trends in large datasets, uncovering hidden connec-

tions often missed in 2D visualizations. Furthermore,

some studies have adopted 3D force-directed graphs,

which not only produce clear visuals but also allow

for interactivity. For instance, a study has shown

that integrating geographic constraints with 3D force-

directed layouts signiﬁcantly enhances the visualiza-

tion of relationships among entities (Wang et al.,

2023). This method enables interactive exploration,

helping users intuitively understand complex patterns

within large datasets. The results of these studies indi-

cate the practicality of using 3D force-directed graphs

for our research. Therefore, we adopt this approach to

visualize the synergies between cards.

3 METHOD

In this study, we employ both text analysis and the

concept of co-occurrence rates as metrics to evaluate

the synergy between cards.

3.1 Synergy Score

First, we provide a brief overview of card character-

istics in DCCGs. Generally, most DCCGs classify

cards into two main types: minion cards and spell

cards. For example, Figure 1 shows a minion card

from the game Hearthstone. A minion card has the

following elements: card name, mana cost, effect,

type, attack points, health points, and rarity. Minion

cards can engage in battles on the game ﬁeld until they

are defeated, and their effects can sometimes persist

beyond a single turn. A spell card, in contrast, does

not have health or attack points. They are typically

single-use cards with immediate effects (Figure 1).

A card’s deﬁning feature is clearly its effect,

which shapes interactions and synergies with other

cards. Therefore, the most effective way to evaluate

card synergy is by analyzing the text of these effects,

as synergy largely depends on them. Therefore, we

would like to calculate a synergy score based on the

effects between two cards.

However, in practice, many card effects in DCCGs

lack strong connections to other cards. The “Faerie

Dragon” mentioned above is a prime example (Fig-

ure 1). For cards with isolated effects, analyzing syn-

ergy through text alone is challenging. In such cases,

Mana Cost

Rarity

Card Name

Rarity

Effect

Type Type

Attack

Point

Health

Point

Figure 1: A minion card and a spell card from “Hearth-

stone”.

we can utilize co-occurrence networks to incorporate

player choices as a supplementary analytical method.

Co-occurrence networks could reveal clusters of cards

that frequently appear together, which suggest that

these cards may have strong synergy with each other.

We use the co-occurrence rate between two cards, cal-

culated based on co-occurrence networks, to assist in

calculating the synergy score.

Overall, the synergy score between two cards c

and c

is calculated based on the following formula:

synergy(c

) = base(c

) · cor(c

)

where base(c

) represents the basic synergy score,

derived from analyzing the effects and parame-

ters of the two cards, and cor(c

) represents

the co-occurrence rate, calculated based on the co-

occurrence rate between the two cards. The product

of these two values yields our ﬁnal synergy score. In

the following sections, we explain how to calculate

these elements.

3.2 Text Analysis

In DCCGs, most cards typically have various pa-

rameters and effects recorded in their text. We ana-

lyze these texts to detect the synergy between cards.

For example, in “Yu-Gi-Oh!”, “Reinforcement of the

Army” (lower in Figure 2) has the effect of adding a

Level 4 or lower Warrior-type monster to the hand,

and “Sky Striker Ace - Raye” (upper in Figure 2)

meets this condition, making it a valid target for the

effect of “Reinforcement of the Army”. Two cards

that can trigger effect interactions will be considered

to have synergy. Moreover, there are various types

of interactions, and the synergy scores we derive will

vary depending on the nature of these interactions.

The synergy score obtained in this way is referred

to as the basic synergy score. The speciﬁc calculation

formula is as follows:

base(c

) = max(w

w · σ

σ(c

),base

min

)

CSV: Visual Support for Understanding Card Synergy in Digital Collectible Card Games

613

Figure 2: A monster card and a spell card from “Yu-Gi-

Oh!”.

where

w =



effect

archetype

material



σ(c

) =



effect

),σ

archetype

material

)]

Vector-valued function σ

σ(c

) represents the

scores for each evaluation criterion, while vector w

represents the weights of criteria. Additionally, we

designed base

min

(= 1 by default), which serves as a

safeguard to prevent the basic synergy score between

cards from falling to zero, particularly ensuring that

the synergy between cards frequently used together is

properly evaluated.

Regarding the evaluation criteria, they are catego-

rized into the following types:

• Effect Score σ

effect

): This is an indicator

function that shows the degree of relevance be-

tween the effects in the text of the two cards. It de-

termines a value based on whether the any card el-

ements (e.g., card name, archetype, attack points,

etc.) is mentioned in the effect text. If the full card

name is mentioned, the value is 2. If the archetype

is mentioned, the value is 1. If other elements are

mentioned, the value is 0.5. Otherwise, it is 0.

• Archetype Score σ

archetype

): A card may

belong to an archetype (also called a series) that

means a group of cards supported each other.

In “Yu-Gi-Oh!”, cards belonging to an archetype

may contain a common string (e.g., “Blue-Eyes

White Dragon” and “Blue-Eyes Abyss Dragon”)

(Yu-Gi-Oh! Wiki, 2024). In Hearthstone, the

archetype is represented by the card type. These

cards usually support each other by their card

effects. If cards c

and c

belong to the same

archetype, the value is 1. Otherwise, it is 0.

• Material Dependency Score σ

material

Sometimes, playing a card requires prerequisites.

If card c

is speciﬁed as a required material for

card c

, where a required material refers to any re-

source or condition needed to play or activate a

card, the value is 1. If c

meets the conditions to

be used as a material but is not strictly required,

the value is 0.5. Otherwise, it is 0.

Additionally, the calculations for σ

effect

and

material

would be performed twice. The ﬁrst calcu-

lation is based on the text of the ﬁrst card, and the

second is based on the text of the second card. Fi-

nally, the two calculated values of such scores would

be summed as the ﬁnal score.

Next, the following explain the weights of evalua-

tion criteria:

• Effect Weight w

effect

(= 3 by default): This rep-

resents the importance when a card is explicitly

referenced in the effect text. Since this element

indicates a direct interaction between cards, it car-

ries the highest weight.

• Archetype Weight w

archetype

(= 2 by default):

This indicates whether the cards belong to the

same archetype. Since the consistency of the

archetype is important in deck construction, a rel-

atively high weight is assigned to this element.

• Material Dependency Weight w

material

(= 1 by

default): This indicates whether the cards could

serve as required materials to play another card

(such as summoning a minion or casting a spell).

Although some cards rely heavily on speciﬁc ma-

terials, this is typically mentioned in the effect

text, or the cards may directly belong to the same

archetype. In most cases, the majority of cards do

not require prerequisites to be played. Therefore,

this weight is assigned a relatively low weight.

3.3 Co-Occurrence Rate

The co-occurrence rate between two cards represents

the frequency at which they appear together in a given

dataset. A high co-occurrence rate may indicate that

players frequently use the two cards together, which

suggests they have complementary effects, form part

of a popular deck archetype, or provide a strong tacti-

cal advantage when combined. Analyzing these rates

could help in understanding card synergies.

In general, we will collect a large dataset of decks

from elite players to analyze the co-occurrence rate

between pairs of cards. We deﬁne “co-occurrence” as

the concurrent presence of two cards within the same

deck. The formal deﬁnition of the co-occurrence rate

is as follows:

Given cards c

,...,c

and decks

,...,D

, where each deck is a multiset

j,1

j,2

,...,c

j,k

of k

cards, the co-

occurrence rate cor(c

) of two cards c

and c

deﬁned as the ratio of the number of decks, where the

HUCAPP 2025 - 9th International Conference on Human Computer Interaction Theory and Applications

614

two cards co-occur to the number of decks that have

at least one of the cards.

cor(c

) = 1 + k ·





| c

∈ D

∧ c

∈ D







| c

∈ D

∨ c

∈ D





In this formula, the adjustment coefﬁcient k (= 1

by default) is used to control the extent to which it

inﬂuences the synergy score. Moreover, in order to

visualize newly designed cards by designers (as unim-

plemented cards cannot have co-occurrence rates cal-

culated), the co-occurrence rate has the minimum

value of 1.

3.4 Visualization

A 3D force-directed graph is ideal for visualizing syn-

ergies between cards in DCCG, as it can effectively

represent complex networks with a large number of

nodes and edges. The layout algorithm uses attractive

and repulsive forces to naturally cluster related cards

and separate unrelated ones, which could provide a

clear view of card relationships. As a result, we repre-

sent cards as nodes and synergy scores as edge lengths

to construct a 3D force-directed graph for visualizing

card synergies.

To simplify the graph, we deﬁne a threshold de-

noted as α with 0 ≤ α ≤ 1 to ﬁlter out unnecessary

information. If the synergy score is higher than this

threshold, it is inferred that a meaningful relationship

exists between the two cards, allowing them to be in-

corporated into our visualization.

In addition, we use the synergy score as a piv-

otal factor in determining the length of an edge con-

necting the cards c

and c

, which is calculated by

δ/synergy(c

), where δ ≥ 0 is a coefﬁcient for the

adjustment of visualization. The shorter the length,

the higher synergy between two cards.

For implementation, we built upon the Force Di-

rected Diagram plugin (Forgin Bits, 2024) in Unity,

modifying it for our purposes. In addition to the de-

fault rotation and zoom features, we have added ﬁlter-

ing functionality (which reduces the opacity of non-

adjacent nodes when a node is selected) and a search

feature (allowing users to jump to the corresponding

node by searching for keywords) to enhance the over-

all user experience.

4 CASE STUDY

To validate our approach, we selected “Yu-Gi-Oh!

Master Duel” (YGOMD) as the primary subject of

our case study. YGOMD is a digital version of “Yu-

Gi-Oh!” (YGO) with the same game rules and card

Figure 3: Overall view of the 3D force-directed graph with

1,700 cards.

“Spright”

Archetype

“Evil★Twin”

Archetype

Figure 4: Partial view of a 3D force-directed graph related

to “Spright” and “Evil Twins” cards.

pool. We chose YGOMD because it lacks the re-

source constraints that other DCCGs such as “Hearth-

stone” and “Shadowverse” impose with mana costs,

which typically force players into multiple rounds of

battles. In YGO, a player’s deck is nearly the entire

resource. The strength of a deck highly depends on

the interactions among the cards, and it is common

for players to end a match in one or two turns by using

multiple card combinations, which makes visualizing

card synergy highly valuable.

To prepare a necessary dataset for our analysis,

we have collected approximately 1,000 decks belong-

ing to elite players from the “Master Duel Meta”

website (Duel Links Meta LLC, 2024) by using a

Python-based web scraper. After calculating the syn-

ergy scores through programs, we build a 3D force-

directed graph to visualize the card network by Unity.

The results are shown in Figures 3 and 4.

Recall that the length of an edge in our visu-

alization represents the synergy between two cards.

Although we did not employ a speciﬁc clustering

method (through the application of physical simula-

tion principles only), we observed that cards belong-

ing to the same archetype naturally formed small clus-

ters. From Figure 4, we could see that “Spright”

cards and “Evil Twins” cards formed clusters respec-

CSV: Visual Support for Understanding Card Synergy in Digital Collectible Card Games

615

tively. In fact, these two archetypes are often com-

bined in a single deck. Through our visualization,

users can better understand the synergy between dif-

ferent archetypes, as those that work well together

are positioned close to each other within our network.

The ﬁltering feature we implemented also makes the

synergy between cards simple and easy to understand.

5 EVALUATION

Given the absence of previous studies with which we

could compare our research, we assessed our study by

conducting a questionnaire survey and a usability test

among DCCG players.

5.1 Questionnaire

5.1.1 Survey Design

We created an online questionnaire with 13 questions

using Google Forms and asked participants to try our

tool. It gathered basic user information, assessed their

understanding of DCCGs as part of the pre-study,

and collected their feedback on our visualization in

the post-study section. The answer scale is primarily

based on a 5-point ordinal scale.

Here are the questions included in the question-

naire, except for the ﬁrst 3 questions that ask for per-

sonal information (name, age, and gender). The fol-

lowing are all required multiple-choice questions:

• Pre-study Part

Q1: How familiar are you with DCCGs?

Q2: Do you think that learning to play a DCCG

requires a lot of time and effort?

Q3: If you have ever played DCCGs, did you ﬁnd

it challenging to build your own deck when you

were a beginner?

Q4: How familiar are you with Yu-Gi-Oh?

• Post-study Part

Q5: What kind of information did you get from

our 3D graph?

Q6: How clearly does the visualization show the

synergies between cards?

Q7: How easy was it to interact with the 3D

force-directed graph?

Q8: Do you think this visualization would help

you in building or optimizing a deck for a

DCCG?

Q9: How accurate do you ﬁnd the results of the

visualization?

Q10: How appealing do you ﬁnd the overall de-

sign of the visualization?

Figure 5: The results of questions Q1 to Q10.

5.1.2 Results

Since our visualization does not provide any informa-

tion on game rules, we require users to have a basic

understanding of DCCGs. Therefore, we surveyed 10

individuals with varying levels of experience in YGO.

Based on the feedback, all participants are males aged

20 to 29. Regarding familiarity with DCCG (not lim-

ited to YGO), the participants included 2 beginners,

2 intermediate players, 3 experienced players, and 3

expert players (Q1 in Figure 5). All had played YGO,

with 2 being beginners, 4 being experienced players,

and 4 being expert players (Q4).

To conﬁrm the presence of information overload,

which leads to high learning costs, we asked partici-

pants if learning to play a DCCG requires signiﬁcant

time and effort. According to the results (Q2), all par-

ticipants answered “Yes”. We also asked whether they

struggled to build their own decks as beginners, and

all participants again responded “Yes” (Q3).

To evaluate our research, participants explored the

interactive 3D graph and identiﬁed the types of in-

formation obtained during the process, with multiple

answers allowed. The results showed that 100% of

participants gained insights into card synergies, the

primary goal of the visualization. Additionally, 50%

learned about archetype synergies, and 20% found in-

spiration for deck building (Q5).

We further assessed the visualization using

multiple-choice questions focused on usability, aes-

thetics, accuracy, practicality, and clarity. For “How

clearly does the visualization show the synergies be-

tween cards?” 40% rated it “very clear”, 20% “mostly

clear”, and 20% “neutral” (Q6). Regarding “How

easy was it to interact with the 3D force-directed

graph?” 80% found it “very easy”, while 20% rated it

“somewhat easy” (Q7). When asked “Does this vi-

sualization help in building or optimizing a DCCG

HUCAPP 2025 - 9th International Conference on Human Computer Interaction Theory and Applications

616

deck?” 20% said “helpful”, 30% “somewhat helpful”,

and 20% were “neutral” (Q8). In terms of accuracy,

for “How accurate do you ﬁnd the visualization re-

sults?” 50% rated it “accurate”, 30% “somewhat ac-

curate”, and 20% “neutral” (Q9). On design appeal,

“How appealing is the overall visualization design?”

received 20% “very appealing”, 50% “moderately ap-

pealing”, and 30% “neutral” (Q10).

5.2 Usability Test

To assess the usability of our research, we conducted

a usability test involving two groups of players: 2

beginners and 2 experienced players. This test con-

sists of 3 sections: Pre-study Questionnaire, Task Per-

formance Assessment, and Post-study Questionnaire.

Since our previous questionnaire was designed to ful-

ﬁll both Pre-study and Post-study functions, partici-

pants have already completed it as outlined in the pre-

vious section, and these responses have been included

in our prior statistical analysis.

5.2.1 Design of Task Performance Assessment

The task performance assessment involves guiding

players through a series of tasks that we have de-

signed, with difﬁculty increasing incrementally. For

each task, we record the completion time, and ob-

serve participants’ approaches to understand their be-

haviors. The assessment includes the following 3 pri-

mary tasks:

T1: Identify the three cards that have good synergy

with “Dark Magician”.

T2: Identify two archetypes that has good synergy

with “Blue-Eyes” Deck.

T3: Use this tool as a reference to create a competi-

tive “Hero” deck in YGOMD.

5.2.2 Results

Table 1 shows the task completion times for each par-

ticipant. For T1, all players used the search bar im-

mediately, completing it in about 10 seconds. Experi-

enced players completed the task ﬂawlessly, while be-

ginners often included staple cards (commonly used

powerful cards) among the top synergy cards instead

of focusing solely on those truly with good synergy

(Yu-Gi-Oh! Wiki, 2024). In T2, beginners took 50

seconds and 1 minute 2 seconds, respectively, while

experienced players ﬁnished in 23 and 26 seconds.

Beginner actions revealed that the visualization does

not clearly convey archetype synergy, which leads to

confusion, despite cards within the same archetype

Table 1: Task completion times for each participant.

Participant T1 T2 T3

Expert 1 12s 23s 5m2s

Expert 2 9s 26s 4m37s

Beginner 1 13s 50s 6m19s

Beginner 2 10s 1m2s 7m44s

forming distinct clusters. For T3, experienced play-

ers took 5 minutes 2 seconds and 4 minutes 37 sec-

onds, while beginners needed 6 minutes 19 seconds

and 7 minutes 44 seconds. Experienced players cre-

ated decks of higher quality, comparable to those in

actual gameplay. In contrast, beginners’ decks were

less organized, often including all related cards with-

out strategic considerations.

The Task Performance Assessment shows experi-

enced players completed tasks faster and with higher

quality due to prior knowledge. Beginners, though

slower and less accurate, still performed reasonably

well with the tool’s assistance.

6 DISCUSSION

6.1 User Feedback Analysis

Based on user feedback, we can conclude that our vi-

sualization effectively helped players understand the

synergies among cards. The survey conﬁrmed that

DCCGs often have a high entry barrier due to infor-

mation overload. For all questions evaluating differ-

ent aspects of the visualization, participants provided

either neutral or positive responses.

Nevertheless, the results also indicate that while

the visualization is relatively clear, simply under-

standing the strength of card synergies provides lim-

ited assistance in deck building and overall game

comprehension. Additionally, the tool’s effectiveness

seems to be inﬂuenced by the player’s experience

level in DCCGs. For example, in task T1, beginner

players who were less familiar with the concept of

staple cards demonstrated relatively lower accuracy in

their responses. Offering clearer delineation of cluster

boundaries could help address this issue.

6.2 Application on Other DCCGs

Our goal is to propose a method for visualizing card

synergies across a wide range of DCCGs. In this re-

search, the visualization experiment was conducted

exclusively with YGOMD. While this demonstrates

that our approach could be effective for games where

cards have strong interactions, additional visualiza-

tion experiments on a variety of DCCGs are planned

CSV: Visual Support for Understanding Card Synergy in Digital Collectible Card Games

617

to further validate the feasibility of our method.

In addition, in games like the “Pok

emon Trad-

ing Card Game”, card texts often lack explicit inter-

actions, which necessitates relying primarily on co-

occurrence rates for synergy score calculations. To

address this limitation, we plan to simulate and col-

lect gameplay data to evaluate how cards interact in

practice, aiming to explore hidden synergies that are

not explicitly described in their texts.

6.3 Reﬂection on the Evaluation

Method

The lack of similar studies on DCCGs presents a

signiﬁcant challenge to conducting objective, cross-

comparative evaluations of this tool. This has resulted

in current evaluations relying heavily on subjective re-

sponses, which are less convincing.

Furthermore, the small and homogeneous sample

size used in the evaluation limits the generalizabil-

ity of the ﬁndings. Since our study does not include

features to introduce the game, participants were re-

quired to have a basic understanding of DCCGs. Re-

lying solely on experienced DCCG players may intro-

duce bias by excluding the perspectives of beginners.

Expanding the participant pool to include a more di-

verse audience and integrating qualitative data is one

of our future plans to strengthen the robustness of the

evaluation.

7 CONCLUSIONS AND FUTURE

WORK

In this paper, we introduced a novel approach to visu-

alizing card synergies to overcome information over-

load in DCCGs through a graph-based visualization

framework. To test it, we visualized card synergies for

YGOMD and conducted a questionnaire survey with

a usability test involving recruited volunteers. Al-

though there are areas that require improvement, par-

ticipant feedback conﬁrmed the effectiveness of our

visualization approach.

Our ultimate goal is to evolve this framework into

a comprehensive support tool, enabling our synergy

analysis method to be applied across most DCCGs.

Based on the results of this study, we will continue re-

ﬁning the framework to beneﬁt both players and game

designers, helping users gain deeper insights into card

synergies and the game itself.

ACKNOWLEDGMENT

This work was partly supported by JSPS KAKENHI

Grant Number JP24K14904.

REFERENCES

Chen, Z., Amato, C., Nguyen, T.-H. D., Cooper, S., Sun, Y.,

and El-Nasr, M. S. (2018). Q-DeckRec: A fast deck

recommendation system for collectible card games. In

Proc. IEEE CIG, pages 1–8.

Dockhorn, A. and Mostaghim, S. (2019). Introducing the

Hearthstone-AI competition. arXiv, 1906.04238.

Duel Links Meta LLC (2024). Top decks. https://www.

masterduelmeta.com/top-decks.

e Silva Vieira, R., Tavares, A. R., and Chaimowicz, L.

(2024). A taxonomy of collectible card games from

a game-playing AI perspective. arXiv, 2410.06299.

Forgin Bits (2024). Force directed diagram. https://www.

youtube.com/watch?v=hWt5WPO8c6A.

Fudolig, M., Alshaabi, T., Arnold, M., Danforth, C., and

Dodds, P. (2022). Sentiment and structure in word co-

occurrence networks on Twitter. Appl. Netw. Sci., 7:9.

Konami Digital Entertainment (2021). Yu-Gi-Oh! Master

Duel has over 10,000 cards to unlock. https://www.

konami.com/games/us/en/topics/1752/.

Puerta, P., Laguna, L., Vidal, L., Ares, G., Fiszman, S.,

and T

arrega, A. (2020). Co-occurrence networks of

Twitter content after manual or automatic processing.

a case-study on “gluten-free”. Food Qual. Prefer.,

86:103993.

Strother, J. B., Ulijn, J. M., and Fazal, Z. (2012). Visualiz-

ing instead of overloading: Exploring the promise and

problems of visual communication to reduce informa-

tion overload. In Information Overload: An Interna-

tional Challenge for Professional Engineers and Tech-

nical Communicators, pages 203–229. Wiley-IEEE

Press.

Swacha, J. (2021). Three dimensions of science: A web tool

for 3d visualization of scientiﬁc literature. In Proc.

ACM/IEEE JCDL, pages 274–277.

Turkay, S. and Adinolf, S. (2018). Understanding on-

line collectible card game players’ motivations: a sur-

vey study with two games. In Proc. OzCHI, page

501–505. ACM Press.

Vieira, R., Tavares, A. R., and Chaimowicz, L. (2020).

Drafting in collectible card games via reinforcement

learning. In Proc. SBGames, pages 54–61. IEEE CPS.

Wang, S., Li, W., and Gu, Z. (2023). GeoGraphViz:

Geographically constrained 3D force-directed graph

for knowledge graph visualization. Trans. GIS,

27(4):931–948.

Yu-Gi-Oh! Wiki (2024). Gaming terms. https:

//yugioh.\linebreak[0]fandom.com/wiki/Category:

Gaming\

Terms.

Zuin, G., Chaimowicz, L., and Veloso, A. (2022). Deep

learning techniques for explainable resource scales

in collectible card games. IEEE Trans. Games,

14(1):46–55.

HUCAPP 2025 - 9th International Conference on Human Computer Interaction Theory and Applications

618