An Iconic Elden Ring Weapon Similarity Analysis
By Dante Caraballo and ChatGPT
Motivations
I was motivated by my love for video games, specifically RPGs like Elden Ring. I thought it would be insightful to analyze the game data of one of my favorite games. Often, within Elden Ring specifically, I have struggled to find new weapons similar to the ones I have used in the past. Additionally, the first query I decided on was the “Greatsword”. This weapon is by a popular anime character (Guts) that inspired the game. This iconic weapon was really what motivated me to do my similarity-based analysis. The other weapons I chose were weapons that I enjoyed using in past Elden Ring playthroughs.
Practicality
Question of Interest: How can measuring the similarity between different types of weapons inform a video game developer about balancing gameplay? Specifically, how can we find weapons most similar to chosen pics to find diverse yet similar weapon options for players?
Stakeholder: The primary stakeholder is a video game development team, particularly those involved in game design and balance. Additionally, another stakeholder could be gamers who are trying to find similar weapons to their favorites in the game.
Decisions Informed: Selecting a variety of weapons for the game that are distinct yet balanced, enhancing player experience by offering multiple viable options without any single weapon being overwhelmingly superior. Patch notes and future updates could be influenced when it comes to balancing decisions. Lots of gamers like to min-max their builds (make them the best possible) this could be a valuable resource when min-maxing Elden Ring builds.
The Data
Dataset Description: The dataset contains attributes for various weapons used in a fictional RPG, including:
· Name and Type of the weapon (e.g., sword, katana, seal).
· Damage attributes: Physical (Phy), Magical (Mag), Fire (Fir), Lightning (Lit), Holy (Hol).
· Critical Strike Chance (Cri), Stamina (Sta) usage, and Weight (Wgt).
Relevance: These attributes directly impact gameplay, affecting how players choose weapons based on their play style, the challenges they face, and the strategies they use to survive within the game. Additionally, it directly determines the amount of damage a player can deal to enemies.
Data Collection: Data was extracted from an existing dataset that was available for download at https://www.kaggle.com/datasets/l3llff/-elden-ring-weapons. I found this data on Kaggle, and it contains most if not all the year-one weapons for Elden Ring.
Measuring Similarity
Similarity Metric: I used the Pandas and SciPy modules to execute my analysis. The Euclidean distance was chosen for its simplicity and effectiveness in quantifying the “distance” between weapons in this multidimensional attribute space.
Features Used: The similarity between weapons was measured using their damage attributes, critical strike chance, stamina usage, and weight, as these are key factors influencing a weapon’s performance and player choice.
Findings and Tables
This analysis identified alternative weapons similar to the “Greatsword,” “Moonveil Katana,” and “Erdtree Seal” within a role-playing video game (Elden Ring), utilizing Euclidean distance to assess similarity based on key attributes. For each query item (weapon), the topmost similar weapons were listed, revealing valuable insights for game design, such as identifying gaps in weapon diversity and ensuring balanced equipment options for players.
For each query item (e.g., “Greatsword”, “Moonveil”, “Erdtree Seal”), the topmost similar weapons were identified based on their similarity scores, demonstrating how players can find alternative equipment matching their preferences. The Erdtree Seal and Moonveil Kitana had less than ten other weapons that were considered similar to them. This was interesting which could suggest that developers may need to add more weapons to the katana and seal weapon types.
Data Cleaning and Bugs
Common Bugs: Potential issues include missing values or inconsistent data (e.g., different scales for damage types). These were addressed by replacing missing values with zeros and normalizing attribute values where necessary.
I started each analysis by addressing the common bugs. Next, I began sorting and taking a subset of the data types that I was looking at for each specific weapon. For example,” katana”, and “seal” are two of the weapon types that I filtered by in my analysis.
Limitations
Limitations: The analysis may not account for unique weapon abilities or player preferences beyond the measured attributes, potentially overlooking factors like scaling to the player’s level. However, it is nearly impossible to account for individual builds and playstyles with this type of analysis due to the scope of the game’s leveling system. Additionally, due to frequent updates more weapons may be available in the game that are not present in the data used for this analysis.
Bias: The selection of attributes and the Euclidean distance metric may prioritize certain weapon characteristics over others, possibly reflecting my bias in what constitutes similarity. Additionally, I decided to search by the specific weapon type. In the analysis, similarity could potentially be defined differently, for example not taking the weapon’s type into account or even more than one weapon type.
Conclusion
This study demonstrates the utility of similarity measurement in identifying alternative equipment options in RPGs, offering insights into game balance and diversity. While effective, the approach emphasizes the need for incorporating broader gameplay elements and player feedback into future analyses to capture the full spectrum of strategic diversity and player preferences. These findings serve as a foundation for game developers to enhance player engagement and strategic options in game design.
