Week 1 · Bootstrap edition
Meet the network: 303 heroes, 17 loners, and one island of Morituri
POW! Below isn't a screenshot of a graph — it's the actual graph. Drag a node, hover a hero, search your favorite, hit the dice for a random one. The plots further down are live too: flip axes, swap in-degree for out-degree, watch curves straighten. Go break something. (You can't — it's read-only. But try anyway.)
What we asked
Week 1 hands every group the same frozen snapshot: 303 Marvel characters from Wikipedia's
Category:Marvel Comics superheroes, with a directed edge A → B whenever A's article
links to B's. Before doing anything fancy with it, we wanted three plain questions answered:
how connected is this thing really, who's the most talked about versus who talks about others
the most, and is it really one single blob — or does it secretly fall apart into pieces?
What we did
We loaded the edge list and node roster with every node added first, so the isolated characters (people whose articles link to nobody, and whom nobody links to) survive instead of quietly vanishing from the graph. That gave us:
The gap between 1,784 and 1,434 isn't a bug — it's reciprocity. 350 character pairs link to each other in both directions (e.g. two Avengers who each mention the other), so those 700 directed edges collapse into 350 shared undirected edges once direction stops mattering; the remaining 1,084 edges are one-way and pass straight through. 350 + 1,084 = 1,434.
Poke the universe
One figure, the whole network: the giant component, the Morituri island, and the 17 loners, all in the same force-directed layout. Node size and color both track in-degree — the brighter and bigger a node, the more other heroes point to it.
Drag to rearrange, scroll/pinch to zoom, click a node (or a chip in its panel) to jump to a neighbor. Click empty space to deselect.
Who's linked to, and who links out
In-degree and out-degree tell very different stories about the same page. In-degree is a proxy for fame — how often other characters' articles bother to mention you. Out-degree is a proxy for how thorough the article is — how many other characters get name-dropped in your own page's "allies," "teams," and "enemies" sections.
| Top 5 by in-degree | |
|---|---|
| Spider-Man | 106 |
| Hulk | 64 |
| Wolverine | 60 |
| Doctor Strange | 50 |
| Deadpool | 33 |
| Top 5 by out-degree | |
|---|---|
| Betsy Braddock | 28 |
| Cloak and Dagger | 24 |
| Adam Warlock | 22 |
| Venom | 21 |
| She-Hulk | 20 |
Almost no overlap between the two lists. Spider-Man is linked to by roughly a third of the entire giant component, but his own page doesn't need to name-drop many other heroes to make its point — it's already the most famous article. Characters like Betsy Braddock, on the other hand, have had decades of tangled team-roster history (X-Men, Excalibur, X-Force...), so their pages spend a lot of words cross-referencing teammates — high out-degree, but nowhere near as famous as an inbound link target. TRY IT search "Spider-Man" above and watch how much of the web lights up around him.
What surprised us: an island of Morituri
We expected the isolates (17 characters nobody links to and who link to nobody) — the course page warned us they were worth a look. What we didn't expect was a second, self-contained cluster: a 9-node clique of characters from Marvel's "Strikeforce: Morituri" series, all densely linked to each other and to nobody else in the whole 303-character set. Toggle it off above and watch it detach cleanly from the rest of the universe — it really is its own little galaxy.
The distributions, live
Same data, reshaped: how many characters have exactly k incoming or outgoing links, on linear vs. log axes, with the course's log-binning scheme overlaid — and, for fun, how the real in-degree distribution stacks up against synthetic exponential and power-law samples.
Plotted against k+1 so isolates (k=0) stay visible on the log axis. The gold line is the Goodies' log-binning scheme — width-1 bins up to k+1=7, doubling after — and it sits exactly on top of the raw dots wherever a bin is still width 1. Hover a raw dot for example characters at that degree.
10,000 samples each from an exponential and a power-law distribution, plus the real Marvel in-degree distribution overlaid. Toggle both axes to log to see which curve goes straight — then judge for yourself which one the real data resembles.
Our read: the exponential straightens out on a linear-x/log-y plot, the power law straightens on log–log, and the real in-degree data sits closer to the power law's line in the tail — though with only 303 nodes against 10,000 synthetic samples each, we're not staking much confidence on that yet. TBD we'll actually fit exponents and test this properly in week 2.
What's next
Next week we start pulling the real text behind these pages apart. For now: check the real-degrees and two-distributions explorables against our numbers above — they should match, since we're all working from the same frozen snapshot.