A weekly interactive data story mining the network of 303 Marvel Comics superheroes on Wikipedia — who links to whom, who's a hub, who's an island. One post a week, one shared dataset, eight weeks of digging.
Weeks
Week 1: Mapping the Marvel network
1 dataset
Week 2: Models & null models
7 datasets
-
303 nodes · 1,784 edges
Marvel Comics superheroes
The original null-model tests: three kinds of "nothing special" against the network that started Week 2's questions.
-
21 nodes · 151 edges
The Silmarillion characters
A smaller world where chance stays viable — reciprocity is the only structure the null models can't explain.
-
50 nodes · 720 edges
A Song of Ice and Fire
Five hubs clear the ER ceiling at once, the shuffle explains the clustering away, and 74% of citations run both ways.
-
57 nodes · 790 edges
Middle-earth (list page)
The top hub isn't even a character — and this is the only network whose clustering survives degree-preserving shuffles.
-
22 nodes · 135 edges
Star Trek: Deep Space Nine
Sisko's hub may be pure chance, a Voyager stowaway sits isolated — and reciprocity is still the one real structure.
-
53 nodes · 299 edges
Street Fighter
The most Marvel-like verdict: Ryu's hub can't be chance, the Price model fits on a knife's edge — and Tifa Lockhart stowaways.
-
370 nodes · 1,136 edges
Marvel Comics supervillains
Hulk's hub crushes every random draw, preferential attachment more than doubles it — and an Alien-franchise android stowaways among 43 isolates.
Week 3: Paths, centrality & cliques
2 datasets
-
277-node giant component · 303 nodes · 1,784 edges
Marvel Comics superheroes
Five centrality measures against a 200-shuffle null, a general Six Degrees explorer, a node-removal fragmentation test, and the network's tightest maximal cliques.
-
323-node giant component · 370 nodes · 1,136 edges
Marvel Comics supervillains
The same Week 3 playbook on the supervillains network: Norman Osborn and Hulk top the broken centralities, Purple Man — not a hub — is the single worst point of failure, plus a Six Degrees explorer, fragmentation playback, clique hunting and a villains save-the-network game.
Week 4: Communities, traditions & backbones
1 dataset
Week 5: Build the corpus yourself
303 pages · 727,440 tokens
Week 6: From counts to meaning
303 characters · 727,440 tokens · live analogy arithmetic
Week 7: Coming soon
—
Week 8: Coming soon
—
The Team
- Svetlana Ananina
- Rúni Klein Hansen