Week 1 / Group exercise 1.8
Fame vs. cataloging:
the structural asymmetry
of the Marvel universe.
A directed link means two different things depending on which end you stand on. We split a Wikipedia hyperlink graph of 303 Marvel characters into who receives attention and who gives it, and asked what changes when you stop averaging the two together.
01 / What we asked
One link,
two meanings.
Every arrow in this graph is a Wikipedia link from one character's article to another. An in-link is an endorsement decided somewhere else, by an editor writing a different page; an out-link is a citation decided here, by whoever wrote this page. Our question: how does passive prominence — the fame of being linked to — differ in structure and scale from the active, effortful work of authoring the links out?
in the snapshot
02 / What we did
Freeze the wiki.
Keep every node.
We loaded a frozen snapshot of Wikipedia's Category:Marvel Comics superheroes and every link between those 303 articles. The edge list alone only knows about characters that appear in a link, so we add all 303 nodes first and the edges second — the 17 characters with no link in either direction stay on the roster instead of quietly disappearing.
one giant component
03 / See the network
Every link
has a story.
Click any node to inspect the character and highlight its immediate connections; arrows show the direction of each Wikipedia link. Hub characters are pulled toward the center, quieter ones toward the rim.
High in-degree hubs — Spider-Man, Hulk, Wolverine — sit near the center of the frame. The 9-character satellite island and the 17 isolates drift toward the outer ring, cut off from the giant component. Search below to jump straight to a character.
04 / Investigate
Pick a
character.
Search the shared playground and inspect who links to your character and who your character links to.
Choose a character
Its connections will appear here.
- Choose a character to explore the network.
- Choose a character to explore the network.
05 / In-links vs. out-links
Fame is a lottery.
Editing is a chore.
In-degree is heavy-tailed: a small number of characters are linked to from hundreds of separate articles, each link decided independently by a different editor. Spider-Man tops the list with 106 in-links. Out-degree tells the opposite story — it is bounded by how many peers one person can reasonably cite while writing a single page. The busiest author, Betsy Braddock, links out to only 28 other characters, close to the practical ceiling for one article.
vs. 28 peak out-degree
| Rank | Top in-degree | kin | Top out-degree | kout |
|---|---|---|---|---|
| — | Loading… | — | Loading… | — |
06 / The satellites
Not every path
leads to Spider-Man.
277 characters (91% of the cast) sit in one giant connected component — ignore the arrows and you can walk from any of them to any other. But 9 characters form a separate island with no path at all to the rest of the universe, and 17 more have no links in either direction. The edge list alone would erase all of them; loading the full node roster first is what keeps them part of the story.
a separate island
07 / What surprised us
Three things
we didn't expect.
Reading the topology closely turned up three details the summary statistics alone don't show.
- No middle ground. Connected components come in exactly three sizes: one giant component of 277, one satellite island of 9, and 17 singletons. Nothing sits in between — no organic sub-team of 3, 5, or 20 characters.
- 41 one-way spectators. 41 characters link out to the rest of the universe but are never linked back. They cite their teammates and rivals, yet no other article ever mentions them — they can see the universe, but the universe doesn't see them.
- Spider-Man's monopoly. At 106 in-links, Spider-Man alone accounts for roughly 35% of all 303 characters in the snapshot, leaving a 42-link gap to second-place Hulk (64).
but are cited by none
08 / Conclusion
Split the arrows.
See both stories.
A directed network hides its most interesting behavior if you collapse it too early. Averaging in- and out-degree together would have buried a real structural asymmetry: being noticed scales with how many independent editors exist, while noticing is capped by what a single editor can write in one sitting. The topology is a fingerprint of that constraint — and keeping the full node roster, isolates and satellite island included, is what let us see it.
Out-links are made.”