THE ROYALTY
MAP

Every centrality figure on this map is an answer to a question somebody chose to ask. Count each credit alike and one person is central; weight credits by episodes and somebody else is; ask instead how completely a performer occupied a single kingdom and the order changes again. Switch the model and watch the map change its mind. That is not instability. It is the finding.

How to read it

Size is betweenness, not importance. A node is drawn large when paths run through it, and betweenness is a savage measure: most of the corpus scores zero under every model, so most of this map is small dots, correctly. A large node is not a beloved performer. It is one the graph cannot route around.

Nearness is a claim, and a limited one. Nothing about a node's position is stored — the layout is solved in your browser each time, by springs that pull joined nodes together and a pressure that pushes everything else apart. So distance on the stage approximates distance through the corpus: two names that land near each other are usually few credits apart. Three things it does not know, and which it would be a mistake to read into it. It ignores how heavy an edge is — a 181-episode lead and a single guest appearance pull exactly alike, so the geometry answers the G₀ question no matter which model is selected above. It ignores size — a node is never drawn toward anything for scoring highly. And a fragment, having no edges to the body, has no springs either: where it sits is arbitrary, and only its separateness is a fact. What is reliable is the lit corridors of a selected node; the rest is a two-dimensional shadow of something that does not fit in two dimensions.

Two dials, and they are easy to confuse. The buttons at the top of the page — G₀, G₁, G₂, G₃ — do not change who is on the map. They change how much a credit is worth. G₀ treats every job the same: one appearance, one connection. G₁ says a hundred episodes should not count the same as one. G₂ asks something else entirely — not how much somebody worked, but how completely one show belonged to them, so a performer in all four episodes of a flop scores exactly what a performer in all two hundred of a hit scores. G₃ uses our own labels — lead, regular, recurring, guest — and is marked as a judgement, because that is what it is. The three beneath them, G₀c, G₀p and G₀w, are not models at all: each deletes a whole kind of connection and runs the sums again, to find out whether a result was only ever true because of that kind.

The three genealogies work the other way round. L1, L2 and L3 do not re-weigh this map; they redraw it. L1 keeps only who acted in what. L2 keeps only who wrote, created and produced. L3 keeps only which shows share a fictional world — a spin-off, a crossover, a character who walked out of one series into another. Different lines mean a different cast of characters in each, so a position in one has nothing to do with a position in another, and only positions are ever quoted. Somebody missing from a map gets a dash, not a low placing: it means the question does not apply to them.

Both are switches on this page. The genealogy control is above the model control because it decides what the model is even talking about: pick L1 and the map redraws with only castings on it, pick L3 and the whole corpus shrinks to fifty-two shows in seventeen separate fictional worlds, which is not a gap but the thing that genealogy was drawn to find. Inside a single genealogy the four models switch off, and that is a rule rather than a missing feature — they are arithmetics over the merged graph and do not survive the trip, so size there is simply how many connections a node holds in the map you are looking at.

So: the models argue about how much things count; the genealogies argue about what counts at all. Every record shows both, one under the other, which is how you find the strange cases — a name that rules one map and is simply absent from the next. One honest caveat: the four models run over everything at once, performers and studios and authorship together, which is exactly the mixing the three genealogies exist to refuse. That is what G₀p is for — pull the studios out, see if the answer holds — and the genealogies are the same worry done properly.

A record is a set of doors. The edges listed in the panel are not a description of the node — they are the other nodes, named. Click one and the map holds that node instead, brings it to the middle of the stage, and opens its own record, which lists the edges it holds in turn. That is how a corridor is walked: not by hunting a thousand dots for a name the panel has already told you, but by following the credit that connects them, which is the only route the graph itself recognises. The stage answers in kind: while a node is held, whatever sits at the far end of each lit corridor is named on the map, in the same cyan the corridor is lit in — as many of them as can be printed without one name landing on another, largest first, so what goes unnamed is what the graph routes around least.

A constellation implies a complete sky, and this one is not complete. Every node the main body cannot reach is named in the list beside the map, and each one flies to. None of them is a mystery: they are kingdoms and houses whose corridors nobody has looked for yet, and a corridor is one credit. The list is short because the last two sweeps closed most of it.

← Sitcom Royalty