For this project, I created a network visualization exploring relationships among characters (Legends) in the game Apex Legends, using data on pick rates, average win rates, character classes, player rank distribution, and co-pick frequency. My goal was to analyze which Legends are often used together and how their popularity and performance vary across gameplay tiers. This project builds on my broader theme of analyzing global CO₂ emissions data, but here shifts to a focused and socially relevant entertainment dataset to investigate decision-making patterns in a competitive environment.
Presenting My Work

The final visualization displays 25 Apex Legends characters as nodes. The edges represent how often two characters are picked together in competitive matches. Visual elements were carefully mapped to key attributes:
- Node Size: Reflects the Legend’s average win rate. Larger circles represent characters with higher success rates.
- Font Size (Node Label): Represents pick rate. Bigger labels mean the character is more commonly chosen by players.
- Font Color: Indicates character class (e.g., Assault in red, Skirmisher in yellow, Support in blue, etc.).
- Node Color Fill: Reflects the most frequent rank of players who use this character (e.g., Silver, Gold, Diamond).
- Edge Thickness & Color: Represent co-pick frequency. Thicker and redder lines show stronger synergies between characters.
The graph clearly highlights that Wraith, Pathfinder, Octane, and Bangalore are central and heavily linked to many others. For instance, Wraith shows thick red edges connecting to Valkyrie, Pathfinder, and Horizon—suggesting strong synergy and meta relevance in team composition. On the other hand, lesser-used characters like Rampart and Mirage are visually peripheral, with fewer or thinner connections.
Contextualizing the Work
The dataset was compiled manually based on recent pick rate and performance statistics from Season 20 of Apex Legends, gathered via public community dashboards (e.g., apexranked.com and tournament recaps) and adjusted for consistency in Gephi. Co-pick frequencies were inferred from match logs and pro-team compositions. The metadata (e.g., class, skill style, rank) was collected from official game sources and player-tier reports.
What sets this project apart from existing Apex visualizations is its multi-dimensional design. Most community graphs or tier lists only show raw popularity or win rate in bar form. Here, I combined five dimensions into one rich, interactive visual structure that represents gameplay strategy, player preference, and team dynamics all at once. This approach is inspired by academic network mapping techniques like those in Multimodal Network Analysis, where multiple attributes are layered to uncover complex patterns.
That said, there are known biases in the data. For instance, pro team compositions are overrepresented relative to casual player trends. Pick rates fluctuate seasonally based on balancing updates. Also, some edge weights (e.g., co-pick frequency) are estimated rather than exact, due to limitations in available raw match data.
Reflecting on the Work
The project went through several iterative steps:
- Data Cleaning & Merging: I organized two spreadsheets: one for node attributes (character stats) and one for edges (pairwise relationships). Initially, Gephi couldn’t properly match the two due to column mismatch. After ensuring consistent identifiers across both files (Legend names as ID in the nodes table), the import was successful.
- Gephi Design Decisions: I used the ForceAtlas2 layout algorithm for spatial positioning. I ran modularity detection to group Legends into functional communities, then customized colors for node fill, edge weight, and labels. Font sizes were scaled using logarithmic transforms to avoid overwhelming visual imbalance.
- Iterative Refinement: At first, I used class-based node coloring only. After peer feedback, I incorporated additional visual cues like win rate scaling and added player avatars as background images in nodes. I also adjusted edge opacity and color scale to better distinguish major synergies.
- Final Touches: I exported the network with a black background and overlaid the Apex Legends logo in Figma for presentation consistency. This transformed the graph into a polished artifact suitable for public or industry-facing audiences.
What I Learned and Future Directions
One key takeaway from this project was how much clarity and impact can be enhanced through iterative visual refinement. After receiving peer feedback, I exported the raw network layout from Gephi and brought it into Adobe Illustrator. There, I made detailed edits—replacing some node labels with clearer typefaces, manually adjusting node spacing, and integrating each character’s portrait into their corresponding node. I also modified the color palette slightly to ensure higher contrast on a black background and aligned all elements with the official Apex Legends aesthetic.
This round of edits greatly improved the visual hierarchy and narrative clarity. Peers responded more positively to the finalized version, noting it felt more like an “infographic” than a raw data dump. Through this, I learned that network analysis isn’t just about layout algorithms—it’s also about storytelling and graphic design.
For next steps, I’d like to:
- Animate seasonal pick rate shifts over time using time-series layers.
- Expand from pairwise co-pick relationships to trio-based synergy clusters to reflect actual team dynamics.
- Test usability with actual Apex players to see if the visual aids how they compose squads or understand character roles.
- Explore exporting this into an interactive web version using D3.js for hover-based info display.
In short, this project taught me how design, data, and gaming culture intersect, and how tools like Gephi and Illustrator can together turn raw relationships into compelling stories.