Interactive player map
Plotly analytics
Selected player profile
Player data and exports
Clustering learning guide
What is player-style clustering?
Clustering is unsupervised learning. It groups similar players without requiring predefined labels. The result can support matchmaking, coaching, retention, and personalised content.
Why does scaling matter?
Features can use different numeric ranges. Scaling prevents a large-valued feature from dominating distance calculations. Compare raw and standardised results to see the difference.
How do algorithms differ?
K-means prefers compact spherical groups. K-medoids resists extreme values. Hierarchical clustering builds nested groups, while DBSCAN discovers dense regions and marks isolated players as noise.
How should metrics be interpreted?
Higher silhouette values usually indicate clearer separation. Lower Davies–Bouldin values are preferred. WCSS falls as cluster count increases, so it should not be judged alone.
Why are clusters not always objectively correct?
Player behaviour can overlap and change over time. Different business goals require different features. A useful segmentation must be stable, explainable, ethical, and connected to a real decision.
Responsible use
Avoid manipulative targeting and unfair exclusions. Explain how segmentation affects players. Review groups for bias, privacy risks, harmful assumptions, and changing behaviour.