Every account of television history is a list of names, and every list is an argument nobody has to defend. The greats. The legends. The argument is smuggled in the ordering. This project takes the ordering away from the anecdote and gives it to a method, then reports what the method says even when the method embarrasses the expectation — which it did, immediately and repeatedly.
The corpus is a hand-built ledger: 194 kingdoms from I Love Lucy in 1951 to the shows still shooting, 794 performers, 306 creators, 199 production houses, and every credit between them sourced one page at a time. It is not a scrape. A figure enters when somebody read the page it came from — and where one was reasoned out instead, from a series total or a season range, it is marked as reasoned and the coverage figure does not count it. 90.9% of the credits carry a read episode count; the rest are blank or flagged, and the page would rather say so than round up.
Three maps of the same country
The corpus was read as one network for a long time before it became clear it was three. They use the same material and answer different questions, and a map that merges them answers none of them.
There is no separate page for each of the three, because each is a map rather than an essay. Every card above opens the Royalty Map drawn as that one genealogy and nothing else — L1 with only castings on it, L2 with only authorship and production, L3 with only shared fiction — so the claim that the three do not agree is a thing you can look at rather than a thing you have to take on trust. Every name you open there shows its position in all three at once, including the names that hold a position in only one.
The layers are never mixed. An actor taking another job is not a diegetic edge, however much the pattern invites it: Frasier Crane leaving Cheers for Frasier is one, because the man in the fiction moved too. That refusal is the whole reason the third map is worth drawing — without it, it collapses back into the first one wearing a costume.
The finding
Rank every name in every layer it belongs to, then ask whether the three orders resemble one another. Between the performer layer and the creative layer, across the 210 names both of them place, the rank correlation is 0.164.
That is a weak relationship, and weak is not the same as none. Knowing where a name sits in one genealogy of American sitcom tells you very little about where it sits in the other — enough to notice, nowhere near enough to predict. The industry that makes the fiction and the labour that carries it are close to being two separate hierarchies, and almost nobody occupies both.
This figure moved, and how it moved is the more useful result. When the corpus held 509 performer credits the same correlation read 0.003 — genuinely nothing — and this page said so. Then the ensembles were filled in, the corpus doubled to 1,048 credits, and the number rose to 0.164. Nothing was corrected; the finding simply had less evidence under it than its precision implied. That is what a coverage figure is for, and it is why no number here is quoted without the model that produced it and the share of the corpus that model could see.
| Name | Performer | Creative | Diegetic |
|---|---|---|---|
| Blake Clark | #1 of 794 | not applicable | not applicable |
| Jim Gaffigan | #25 of 794 | #226 of 304 | not applicable |
| Cheers | #11 of 194 | #28 of 193 | #1 of 44 |
Blake Clark is the sovereign of the performer layer, and most people reading this have never deliberately watched him. He is a working character actor — Harry the hardware man on Home Improvement, one episode here, two there, across fourteen shows and four decades. He is not central because he was beloved. He is central because he kept turning up, and turning up is what connects a network.
He wrote nothing. He created nothing. He spawned nothing. In the other two genealogies he is not ranked low — he is absent, which is a stronger and stranger fact. The instrument was built to be able to say not applicable rather than four hundredth, because asking where a character actor ranks among authors is a category error and a number would disguise it as a poor showing.
Jim Gaffigan is the widest disagreement in the corpus. Twenty-fifth of 794 as a performer and two hundred and twenty-sixth of 304 as an author — seventy-one percentile points apart, on a man who has his own show with his own name on it. And Cheers is the rarest shape of all — the one kingdom that radiates in all three at once: a dense ensemble, a foundational authorship, and a fiction that outlived its own title.
Four models, no winner
The headline result depends entirely on how a credit is weighed, so the project refuses to weigh it once. Four graphs are computed over the same edge list and all four are reported: presence, where an appearance is an appearance; episode-weighted, where a hundred episodes is not one; saturation, which asks how completely a performer occupied a single kingdom; and editorial role, which uses our own vocabulary and is labelled interpretive so that nobody mistakes it for a measurement.
They disagree, and they disagree completely. Blake Clark leads the first. Betty White leads both middle two. Ted Danson leads the fourth. Nobody is top-five under all four — every current sovereign is the sovereign of one particular weighting, and the page will not pick one for you. A finding is worth as much as the number of models it survives, so no figure here is ever quoted without saying which model produced it and how much of the corpus that model could see.
Four further graphs exist only to attack the result. One removes every connection the project assigned by hand, because an edge awarded to a wanderer for wandering and then rediscovered by our own arithmetic is circular. One removes the studios and showrunners, because a production house that touched forty shows is a hub by construction rather than by significance. One removes every credit added by the cast-completion research, which is half the corpus, because a finding that rests on one body of work is a claim about that work.
The fourth was built because twenty-seven credits reordered the top five. A guest sweep went down the capture index after the credits a main-cast sweep is designed to skip — one episode here, two there, on performers the corpus already held — and its findings moved the presence ranking further than the cast-completion research had moved it with fifty times as many rows. Betty White went from eleventh to second, Cloris Leachman and Valerie Harper entered the top five, and the sovereign's own score fell by a sixth. That is the measure behaving correctly: a guest credit is a short edge between two distant parts of the corpus, which is precisely the shape betweenness rewards, and it is the reason this project's central finding is about a character actor rather than a star. It is also, still, a claim about one afternoon's research — nine of the twenty-seven carry no episode count because the researcher looked and could not find one — so it comes with the graph that removes it.
Blake Clark does not move under any of the four. He is first under all of them, with a shift of zero. He has held first place through the addition of every institution in the corpus, every episode count in it, a whole layer of directors that did not exist when he first took it, and then the doubling of the corpus by the one correction designed to unseat him. Others are not so steady: Lisa Kudrow sits tenth on the full graph and sixty-fourth without the cast research. One of those is a finding about television. The other is a finding about how the ledger was built.
All four are drawn, and the switch between them is public. The Royalty Map puts the corpus on one stage with the model as a control: change the question and the sizes change with it, because size there is betweenness under whichever model is selected. Every name opens a record: what the ledger knows about it — years, network, episode count and how we came by that count, any figure two sources disagree about, and the note somebody wrote while working it out — then every credit it holds, all four scores with the position each implies, what happens when a whole kind of connection is taken away, and where it stands in each of the three genealogies. Every credit in that record is itself a door to the name at its other end, and every name has its own address: the map opened on Cheers is a link you can send. The disagreement described above is a thing to read here and a thing to operate there.
Ignition, which is not labour
Directing episodes is not an edge here and never has been. James Burrows directed seventy-five episodes of Taxi alone; admit that and the creative genealogy becomes a labour census in which the busiest hand wins by volume. But directing the first episode is a different claim. Whoever shot the pilot set the camera language, the blocking and the room the series then repeated for its entire run — and there is at most one per kingdom, so the volume problem cannot arise. It is the single directing credit the project admits.
All 194 kingdoms now name one. 102 people directed them, and the distribution is the finding: James Burrows has 25, Jay Sandrich 10, Andy Ackerman 7, and a long tail of directors with one apiece. Ignition is concentrated in a few pairs of hands across four decades — but nothing like as concentrated as the anecdote about Burrows says, and the anecdote was written down in advance precisely so the ledger could disagree with it.
Every one of these credits carries a grade for its own evidence. 91 of the 197 cite a specific page; 106 are marked patterned — consistent with what is known, not separately sourced — and the ledger says which is which rather than rounding the second kind up into the first.
A fourth layer, and it will not let you quote it
There is a fourth map under construction: form. Not who circulated or who built it, but which device travelled — direct address, the talking-head interview, the bottle episode, the live broadcast. It is the first layer with no external source of truth. Every other layer rests on a credit, which is a documentary fact with a page behind it. A device is assigned, and a densely assigned graph measures the assigner rather than the corpus.
So four rules were frozen before a single edge existed: a closed vocabulary of twelve devices, each carrying a clause saying what it is not; a density ceiling, because a device shared by forty shows is a genre and not a transmission; an evidence ladder on which resemblance never becomes an edge; and an ablation that drops the heaviest device to see whether the layer's shape survives losing it. The layer runs those tests against itself on every build and publishes the verdict, whatever it is.
It currently passes the ablation and falls short on size — 8 edges against a floor of 12 — so it is excluded from everything above until it holds enough. What promotes an edge is a person, a place and something close to their words: Shandling naming Burns in a recorded interview is a transmission; two shows that merely resemble each other are not. Where a claim is checked and no such statement exists, it is marked retired rather than deleted, so the absence is on the record and the same resemblance is not proposed again by the next person who notices it.
What the ledger refuses to do
An unknown stays unknown. No episode count is ever derived from a series total; no series total is ever derived from a count. A figure with no source is worth less than a blank, and a blank with nothing scheduled against it is not discipline — it is absence, so every gap is a queue with an order of work behind it.
Two people with one name are two people — the Christopher Lloyd who co-created Modern Family is not the Christopher Lloyd you are thinking of, and matching on the name would have run a spectacular false corridor between Taxi and Modern Family. One person with two names is one person, declared and never quietly merged, because a genealogy that silently fuses two humans is worse than one that visibly splits one.
Every state of the corpus is fingerprinted and kept. Nothing is dated by the clock — only by the record as it stood — so a figure cannot change because a script was run on a Tuesday.
The apparatus is a CSV ledger, a zero-dependency builder, a standing integrity report that lists what the project does not know, and an append-only census of every state the graph has passed through. The ledger is plain text and opens in any spreadsheet, so a human and a machine edit the same file and the record shows whose hand made which row.