Ownership Topology Dataset
A graph of game nodes and shared-ownership edges, derived from real player libraries across five publishers.
Overview
An ownership-topology graph: game nodes connected by shared-owner edges, with derived graph-centrality scores and organically detected clusters.
This sample is a curated five-publisher cut. It demonstrates the structure — it is not full-market coverage.
Game nodes
Overlap edges
Ownership edges
Organic clusters
Structure
Each node is a game carrying ownership counts and derived graph-centrality scores.
| Field | Type | Description | Example |
|---|---|---|---|
| id | string | Internal game identifier | 73072 |
| title | string | Game title | Surviving Mars |
| publisher | string | Publishing company | Paradox Interactive |
| sample_owner_count | int | Players in the sample who own this game | 25 |
| connectivity_score | float | Normalized count of overlap edges (0.0 to 1.0) | 0.93 |
| proximity_score | float | Normalized closeness to other games in the graph | 0.94 |
| authority_score | float | Normalized influence within the game's cluster | 1.00 |
| cluster_id | string | Organic cluster this game belongs to | Cluster 3 |
| steam_app_id | string | Steam application identifier | 464920 |
Relationships
Each edge connects two games co-owned by players in the sample.
| Field | Type | Description |
|---|---|---|
| source_game_id | string | First game in the co-owned pair |
| target_game_id | string | Second game in the co-owned pair |
| shared_sample_owners | int | Players in the sample who own both games |
| jaccard_sample | float | Jaccard similarity of the two owner sets |
| same_publisher | bool | Whether both games share a publisher |
Preview
Representative game nodes from the sample, ranked by owner count.
| id | title | publisher | sample_owner_count | connectivity_score | cluster_id |
|---|---|---|---|---|---|
| 73072 | Surviving Mars | Paradox Interactive | 25 | 1.00 | Cluster 3 |
| 78458 | Cities: Skylines | Paradox Interactive | 24 | 0.93 | Cluster 3 |
| 26524 | Golf With Your Friends | Team17 | 20 | 0.72 | Cluster 1 |
| 73862 | Dredge | Team17 | 20 | 0.71 | Cluster 1 |
| 73088 | Overcooked! 2 | Team17 | 20 | 0.76 | Cluster 2 |
| 77866 | Blasphemous | Team17 | 19 | 0.68 | Cluster 1 |
| 84466 | Crusader Kings II | Paradox Interactive | 19 | 0.80 | Cluster 1 |
| 4338 | Automachef | Team17 | 17 | 0.64 | Cluster 2 |
Definitions
Graph-centrality measures computed from the overlap topology.
| Signal | Range | What It Measures | Interpretation |
|---|---|---|---|
| connectivity_score | 0.0 to 1.0 | Overlap-edge count per node | How many other games share owners with this one. |
| proximity_score | 0.0 to 1.0 | Graph closeness centrality | How central a game is across the whole topology. |
| authority_score | 0.0 to 1.0 | Within-cluster influence | How strongly a game anchors its own cluster. |
| jaccard_sample | 0.0 to 1.0 | Owner-set overlap ratio | Edge strength independent of raw popularity. |
Methodology
Player libraries are intersected pairwise across games. Every pair of games co-owned by enough sampled players becomes a weighted overlap edge.
Clusters are detected from the edge topology alone — community detection on the overlap graph, with no genre or publisher labels supplied as input.
Connectivity, proximity, and authority scores are normalized to a 0.0–1.0 range within the sample, so nodes stay comparable regardless of absolute popularity.
FAQ
No. It is a curated five-publisher sample chosen to demonstrate the topology. Production datasets cover the full catalog.
Purely from shared-ownership structure. Community detection runs on the overlap graph; genre labels are applied afterwards, never as input.
An edge means players in the sample own both games. Its weight is the count of shared owners, normalized by Jaccard similarity.
Yes. Game nodes carry Steam application IDs, so the graph joins cleanly onto first-party catalog and ownership data.
This sample is the surface of the ownership graph.
The full dataset maps the publisher ecosystem at scale.