sw-series-structure

Series Structure: Episodes and Seasons (剧集结构：单集与季)

Output language = question language; terms are always anchored back to the original word, see sw-workflow/terms.md, do not invent translated names.

Division of labor:Calvisi gives the page-range for each beat of the pilot, Douglas gives the four-act grid and scene counts, Oberg gives the foundation of "format vs. structure" and the four tools of information management, Landau gives act-count sampling and cliffhanger typology, Goldberg & Rabkin give the function table of the classic broadcast four-act, Miller gives the landmark and the five-piece scene structure set, Rabkin gives the speed-up rules of the streaming era, Blum gives the act-count and page-count table for each program type, and the five script collections (West Wing / Succession / Sopranos / Downton / Fleabag) give verifiable measured numbers.

For tables, the complete beat sheet, and episode-by-episode data, see the same directory reference.md。

For the four-column comparison table of the English terms used in this pack, the official translated names from Yan Min's Chinese translation, and one-sentence definitions, see reference.md Section 12 (including the two odd translations log line＝"计程绳" and staff writer＝"试用编剧", as well as the warning about "晚进早出" and enter late, leave early being the same name with opposite meanings).

Not in this skill: series engine / franchise / bible / pilot types and how to sell them → sw-series-engine-bible; jokes and format of the half-hour comedy → sw-sitcom-comedy; writers' room process and document chain → sw-writers-room; episode counts and episode system of domestic Chinese dramas → sw-chinese-series-practice; full text of the episode-by-episode structure table → sw-series-case-studies / succession-series-writing. Medium-agnostic craft is not rewritten:for scene value turns see sw-scene-craft, for dialogue see sw-dialogue, for the three-act paradigm and plot points see sw-story-structure。

Which sections to read by task (do not read straight through)

This section is 38 KB; reading it section by section is usually a waste. Read only the corresponding few sections according to the task at hand:

Task	Read	Skip
Break the structure of one episode / lay out acts	Two (container and acts), Three (act-outs and information management), Four (A/B/C weaving)	Five, Seven, Eight
Write or revise a pilot	Five (pilot beat sheet), Two	Seven (season shape), Six
Diagnose an existing outline or script	Nine (diagnostic checklist), then go back to the corresponding chapters according to the failed items	all the rest
Lay out an entire season	Seven (season shape), Four	Five, Six
Write a half-hour comedy	only the medium-agnostic parts of Three and Six → go to `sw-sitcom-comedy`	the page numbers of Two and Five (one-hour only)
Write a domestic Chinese 45-minute drama	Three, Four, Six → for time slots and scene counts go to `sw-chinese-series-practice`	the page numbers of Two and Five

reference.md is for looking things up; open it only when SKILL.md explicitly points to a certain section.

One, the fundamental difference between series structure and film structure

A tighter form, so it is harder, not easier. Douglas, Chinese translation Chapter 3 (p0088): "A TV script requires condensing all the same elements into a more compact form." A film scene can be 5–7 pages; televisiona scene targets 2 pages——"If you write a meeting scene on 4 pages from the start, that's waving a red flag."
A script simulates the finished film after editing, not the shooting material, and it should give the reader the experience of "an audience watching a show" (Calvisi).
The screenwriter personally labels the acts; there is nowhere to hide(Calvisi lists this as the first formatting difference of TV scripts): "There is no place to hide." A pilot without page breaks looks like cheating, and along with shrinking the margins, it is a technique that immediately exposes an amateur.
Scene numbers are not dramatic scenes. Douglas: an exterior of a building "is not a scene in the fully dramatic sense";a dramatic beat can contain multiple slug lines; it syncs only with the development of the story, not with the locations. Oberg's counting rule:protagonist or goal changes → new dramatic scene; location or time changes → new logistical scene; structural documents count by dramatic scene.
Finished runtime and script length are two different numbers(Calvisi): a broadcast network's main program minus commercials is only 45–48 minutes, but a pilot script must be at least 50 pages—better long than short, "but try not to exceed 60 pages."
A series' "acts" are not a dramatic principle; they are ad slots. In Chapter 1, Oberg calls any prescribed division into a fixed number of acts and pages story format (logistical), whose origins are theater changing candles, film changing reels, and TV inserting commercials, "But all this is gone."; the "string of sausages" metaphor: the logistical method only determines "how many sausages and how long."The most important judgment in this skill: the number of acts is a delivery format, while three acts (before / during / after a main dramatic action or evolution) are a design tool.
But structure is also production scheduling. Miller, Chapter 9: the first thing the production department does when it gets a script is count days in and out;write only DAY or NIGHT——if you write dawn, production will schedule shooting precisely for that day's sunrise period.
Structure is inherited. A series follows the pilot's act structure and page ranges (for a comparison of the act pages of the Breaking Bad pilot and the same show's #308, see reference.md Section 1)—writing a pilot is equivalent to setting the structural mold for the entire series.

II. The Single-Episode Container: Teaser, Number of Acts, and Page Anchors

2.1 The Functional Chain of the Broadcast Four Acts (Goldberg & Rabkin, 2003)

Act One Introduce characters, conflict, stakes → The end of Act Two must push the story toward a "startlingly new and unexpected direction"(Law & Order: the police arrest an unexpected person) → At the end of Act Three, the protagonist thinks they are in control but discovers they are wrong，"They are all going to die. There is no hope." → Act Four Come up with a solution and achieve the goal.

Time-sensitive flag (2003): the book says "one-hour shows are always four acts,"now outdated——since 2006, broadcast networks have generally used five acts, ABC's new shows are always six acts (Douglas, Chinese translation p0091), and streaming has abolished act markers. Butthe four-act functional chain is still valid today: both Miller and Landau (2021) still require breaking in four acts first before cutting further.

For the four page numbers of the page anchors (Douglas, Chinese translation p0090–0091, a four-act first draft including a teaser) and the reason behind "a commercial every 13 to 15 minutes," see reference.md Section 2.

⚠️ These page numbers are only valid within the 48–63 page American one-hour episode and cannot be proportionally converted to domestic 45-minute episodes. The two cutting methods are based on different criteria: American episodes cut acts bypage count and ad slotswhile domestic episodes cut bytime slots and segments(main content about 41.5 minutes: 3-minute opening + a small hook every 15 minutes + a big cliffhanger at the end of the episode). When writing domestic dramas, this section only takes the medium-independent parts (act-break typology, an act ending as "the protagonist is pushed forward with no way back," the Act Two break changing the nature of the story, the five-piece scene structure, the stakes test, and the four tools of information management), and the time slots and scene counts should be changed to sw-chinese-series-practice。 Half-hour dramas are likewise not applicable(a 30-minute script is about 30 pages, and both the number of acts and the number of scenes are a different system): for act length, page numbers, and scene counts, see sw-sitcom-comedy Section 4,do not scale the one-hour numbers from this section over to them. The two diagnostics 4 and 5 below in this section are only valid for American one-hour episodes.

2.2 Scene count: 5–7 scenes per act, about 28 scenes per episode

Douglas's arithmetic, the totals of 28 scenes for four acts / 20 scenes for five or six acts, Oberg's A/B/C scene-count ratios, and Chase's "about 35 beats per episode" see reference.md Section 2.

2.3 How to change four acts → five acts / six acts

The three conversion methods of Douglas, Peter Blake (House), and Miller, as well as the six-act tail design in which "Act Six is actually the next episode's teaser," see reference.md Section 2.

2.4 Teaser / cold open

Length: for the teaser length figures of Douglas, Landau, and Oberg, see reference.md Section 3.
Ten types(Douglas's seven paradigms + Landau's three types): for their names and examples, see reference.md Section 3.
Hard rules for beginners(Douglas, Chinese translation p0098): "Insist on opening with story rather than information... many students... mistakenly believe that opening with a major thematic idea or abstract philosophy will make a deep impression.They are actually shrinking from the challenge of seeing the dramatic arc advance. The best hooks tend toward the best dramatization."
Taboos of the teaser flashback(Oberg): you can begin narrating from the middle or even the end of the dramatic action (The Walking Dead, Breaking Bad),but you must never answer the dramatic question in the teaser; if drawing from the ending, choose the climax before or during(Run All Night), not after (John Wick).
The teaser is not essential: for Landau 2e and Oberg's arguments, and the list of shows with no teaser and with "a short first act serving as the teaser," see reference.md Section 3.
Domestic 45-minute episodes: the teaser and the "3-minute opening hook" (sw-chinese-series-practice 5.1) occupy the same position,choose one of the two, and by default keep the hook.
cold open ≠ teaser(Miller): the cold open of the Brooklyn Nine-Nine pilot does not affect the main line and only launches the theme of that episode, so that episode's point of attack falls in the first act.

2.5 Invisible acts: threefold criteria

Paid cable/streaming scripts often do not mark acts or paginate (Calvisi names True Detective, House of Cards, and Mad Men as not marking acts, but Mad Men places FADE OUT at conventional act breaks), however "a pilot script is almost always going to be structured in the conventional 4-6 act template". The triple criteria measured act by act from the Succession scripts:

Triple criteria: ① Story day / time block markers——A change of day is a change of act；② Switches of location and mode of transport；③ The moment the core question turns. Decision standard:Every act must end with a transfer of information or power, and need not have a suspense hook. For the item-by-item Succession tests (S3E9, S4E01, S4E05, S4E08, S4E10), see reference.md Section 4.

Supporting pattern (Succession S4, ten episodes):8/10 episodes compress the whole episode into a container with a fixed program，Program nodes replace act breaks——i.e. streaming writers are still writing acts, they just disguise acts as a schedule. For the container and scene count of each episode, see reference.md Section 4.

2.6 Whether to write ACT on the page—a disagreement that must be preserved side by side

Five positions(Landau 2e's "story over format", the showrunner who wants to see the labels taking the opposite view, Miller's market judgment, Mazzara's act changes on TWD, Alex Pina's milestone moments) see reference.md Section 2.
Decision standard: writing for broadcast network / basic cable / AVOD → mark acts (AVOD, like traditional broadcast networks, does not allow skipping ads); writing for premium / SVOD → break by four acts, delete the labels before delivery or switch to story day markers; if unsure, follow Miller's criterion and see which outlet the content can go to.

2.7 Page count and episode length

Calvisi / Landau / Miller / Blum's page count ranges for pilots and single episodes, Oberg's episode lengths by platform, Landau 2e's trend judgment, and the timeliness flags for each book's year, see reference.md Section 2.

3. Act out and cliffhanger typology, four tools of information management

3.1 Definition, counterexamples, and criteria

Calvisi: every act (ideally every scene) must make people want to know what happens next; the end of an act is often called "Act-Out"。
Landau 2e Chapter 12："A cliffhanger is a plot device that leaves one or more characters facing either a dangerous situation or a shocking revelation."
Oberg：cliffhanger"simply an unresolved conflict"——positions are not limited to the end of an episode or season, but also include "before leaving a strand"。
Counterexample schmuck bait(Landau 2e): the ECG flatlines, the nurse shouts "Code blue!" (ad break) → "I'm getting a pulse".
Today's rule (the most core one in Landau 2e)："forcing a cliffhanger at the end of each act break is deadly… Instead, act breaks can land on a turning point that emerges from characters, not merely from plot."
Criteria (repeated in both 1e Tip 19 and 2e Chapter 12): a cliffhanger must "grow out of character jeopardy, risk, or fear"; a bad cliffhanger is just a plot point that exists for shock value; "even a 'surprise' cliffhanger needs to be (subtly) set up so that the audience feels they should have seen it coming—but didn't."
Position and time limit: one-hour dramas save the climactic cliffhanger for the end of the episode, multi-camera sitcoms usually place it at the midpoint (the ending must press the reset key to return to normal); you can delay revealing the whole truth,but be sure to reveal it before the end of the season(for the original wording see reference.md Section 5).

3.2 Ten kinds of cliffhanger (Landau)

The item-by-item definitions and examples of the 2e seven types (character in peril | ticking bomb | love is in the balance | surprise twist / aha moment | death of a character | discovery of a secret | life events) and the three additional types listed in 1e (lack of closure | mini-cliffhangers | comedy cliffhangers), as well as the two most transferable practices ("complete the reversal with a single word," "do not go out on the laugh"), see reference.md Section 5.

3.3 The eight forms of act out and three general laws

The eight types tested across six episodes of West Wing:suspenseful question / reversal line / emotional revelation / dialogue-free action / prop freeze / semantic reversal of a repeated line / truncated setup / gesture that pays off a plant. Accompanying rules:hand each act out to a different thread；use dialogue-free action or a prop for at least one act out; do not use dialogue for all four acts—a dialogue-free act out must first teach its grammar. For the eight types with examples, the teaching method of Two Cathedrals, and the four types in Downton, see reference.md Section 5.

An act out can be purely an information event. The two episodes in Goldberg & Rabkin's appendix (Martial Law and seaQuest 2032 "Depths of Deceit") share three general laws: **① The end of an act is not the greatest crisis, but "the protagonist being pushed into a situation with no way back"; ② The Act Two break must change the nature of the story, not merely raise the danger; ③ The end of Act Three must turn the protagonist's victory into a new defeat.** For the act-by-act comparison of the two episodes, see reference.md Section 5.

3.4 The components of suspense and the stakes test

Douglas's rhythm formula(Chinese translation p0139; English edition Anticipation — Expectation — Surprise):anticipation → expectation → surprise, "then begin a new suspense segment"; for the beat-by-beat demonstration of the six-beat cold open of NYPD Blue "Hearts and Souls," see reference.md Section 6.
Landau 2e：Suspense ＝ Anticipation + Surprise, preferably unexpected, inevitable but not predictable.
The stakes test (the most practical one): ask yourself whether "doing nothing in the face of the crisis" is a viable option for this character—"If your character is not compelled to solve a problem by the potential of losing something of value, then your stakes are insufficient."
Polarization model: in every episode, every scene, and every act, identify the positive/negative (+/−) charge; the difference from the old approach is thatthe direction of the act break can be positive or negative——"some acts will end on a win"。

3.5 The four tools of information management (Oberg, Chapters 1–2)

Three-dimensional model: a character is a point; add main dramatic action (conscious want) → a line; add main dramatic evolution (unconscious need) → a plane;add managing information → three-dimensional. Managing conflict answers "Who wants or needs what and why?"; managing information answers "Who knows what and when?"

The definitions and usage table of the four tools (dramatic irony / surprise / mystery / suspense) are in reference.md Section 6.

The item-by-item elaboration of the four red lines (① Do not let the protagonist know more than the audience for a long time; ② For a cliffhanger, prioritize an "information bomb"; ③ closed mystery vs open mystery; ④ "you can have some mystery over the nature of the antagonist, but not over the reality of the danger it represents.") is in reference.md Section 6.

Landau 2e's stitching sentence: "it's not only Who Does What When, but also Who Knows What When."

3.6 Speeding up: act out does not mean a fake reversal every seven or eight minutes

For Rabkin's comparison of 24 and Scandal, the exact wording of "move the coolest thing at the end of Act Four to the end of the teaser," the L+3 ratings logic, and "streaming can cut three things," see reference.md Section 5.

IV. The Weaving and Proportions of A/B/C/runner

4.1 Definitions and Ordering

The definitions and ordering of the four (Douglas's "three separate stories" and runner / Calvisi's "letters ordered by amount of screen time" and the three elements of story engine / Landau's A = franchise, B = personal story, runners = C/D / Oberg's "at least A+B per episode, at most 4–5 threads, and only A and B need to be designed as three acts") see reference.md Section 7.

4.2 Hard Rules of Weaving

Franchise (A story) runs through every act; B story and runner are inserted as needed, each act generally has 4–5 beats, ideally ending on the A story and landing on the protagonist's dilemma（Landau）。
Very rarely end an act on the C story—"If the story is worthy of an act break, then it's not your C story – it's probably more your A or B story." (Landau 1e/2e wording is identical; Oberg says the same.)
But the act-ending cliffhanger is not necessarily on the A story(Douglas Chinese translation p0095: "It may be the turning point of the B story"). West Wing pilot actual test: both the teaser and the Act Three act-out are carried by the B story Sam–Laurie.
There is no need for checkerboard alternation; follow the "line of interest." Douglas uses NYPD Blue "Simone Says" Scene 8 (the old woman downstairs killed by a stray bullet) to demonstrate: "From the point of view of the order of plot development on the 'checkerboard' it is not needed, but it is crucial for following the audience's line of interest."
Inserting another thread can create "ellipsis"—the impression that time has passed, letting the same group of characters reappear in a way they "actually could not have experienced that quickly."
No scene is written solely for exposition. "If you need to convey facts, then put the facts in an emotionally charged context."(Douglas Chinese translation p0119)
When planning, keep the threads separate; only weave them when writing the teleplay(Douglas's order for beginners, the opposite of the order she herself uses when analyzing the four-layer simultaneous staging of "Hearts and Souls").
Thread intersections are most effective when there is a thematic link(Landau 1e), family is the strongest link;a runner can grow into a main story；a serial does not have to write about every character in every episode; for examples of each of the three (Terriers' three-level promotion, Game of Thrones' rotation, Mad Men S6E11 Peggy's rat and elliptical storytelling) see reference.md Section 7.

4.3 The A/B dual track of procedural (Ann Donahue, CSI: Miami, Douglas Chapter 3)

Ann Donahue's A = investigation / B = emotion dual track, "warm body by the end of Act One, within the first 17 pages," and the three writers' room orders of The Good Wife / CSI / House see reference.md Section 7.

4.4 Two Alternative Story-Breaking Methods (Must Be Listed Alongside A/B/C)

Wendy West（Dexter）: when broadcast has no acts, the series actually falls back to three acts; "We break the beats by character and then do a weave."On the whiteboard every character has a beginning, middle, and end,do not break by act, break by character, break the protagonist's thread first—"The reversals tend to come where you'd expect — page 40 to 45 out of a 55-page script."
Glen Mazzara（The Walking Dead）：Explicitly rejects A/B/C— Treat Rick as A and the other person as B, then interweave them; "That doesn't always equal a theme. It just pushes the ball further on each one."; then ask "What is this episode about?"(the episode where the girl walks out of the barn, the episode where the herd breaks through the corral—every character has their own place in that larger story).He himself stated the cost: secondary characters may forever remain secondary characters.
Oberg's compromise: multiple storylines do not equal theme-led—if the characters in each storylineshare the same external problem and the same goal, it is a plot-led co-protagonist structure (in Stranger Things, A1 Joyce / A2 Hopper / A3 the boys / A4 Jonathan are actually all the same A storyline).
Downton's extreme value (Fellowes): the ensemble-drama specification of running 8–12 storylines simultaneously in S2 and "at least one storyline per episode unrelated to that episode's main crisis" see reference.md Section 7.

4.5 Using grid analysis and planning (Douglas's grid, translated into Chinese as "frame grid")

The basic four-act grid, the three steps of breaking a story, reverse-engineering, the requirements of a log line, and Oberg's three-step alternative method (Mapping → Sequencing → Weaving the Strands) see reference.md Section 7.

5. Pilot beat sheet

Calvisi's Benchmark page ranges, the complete beat chain, the verification numbers for AOP and DOW, the four "televisual" exclusive beats, the boundary of "This is form, not formula.", Miller's six-landmark model, and Landau's three strategies for the pilot ending see reference.md Section 1.

The pilot's most critical trade-off (Calvisi): deciding "how many characters and how many storylines to introduce"—cram in too many → too dense; focus on too few → readers do not feel a big story that can sustain 100 episodes.The ending should be good, but it cannot cash out everything, otherwise it will "feel too complete or 'closed-ended'—like a feature"; it must "leave us hanging to some degree and wanting more".

The choice of premise pilot / typical-episode pilot / hybrid, the three ways to end a pilot, the bible and pitch documents → sw-series-engine-bible。

6. The five-piece scene structure and the scene-level checklist

The five-piece set(Miller, Chapter 9, adapted from Lance Gentile, writer on ER / Third Watch and emergency physician):hero's want / obstacle / escalations / decision / resolution。

want and obstacle are already established before the opening；the first escalation opens the scene；decision = the scene's climax；resolution = the result of the decision, and carries the character into the next scene.
Quantity: at least three escalations, "but it's better to have more. More escalations give the scene time to build and breathe."Prefer action——"If all of your escalations are oral… Your TV show turns into a radio show." (the measured escalations of the two scenes in Hannibal and HTGAWM see reference.md Section 8.)
A scene has only one hero, and the hero can change(the men's room scene is Will, the next Jack–Alana scene is Jack).One-person scene: the opposing want comes from the world or the environment—in OITNB "The Chickening", Piper's want is to enjoy her Sunday morning ritual,the obstacle is the prison。
moment / image ≠ scene："If there isn't a strong character want, it's not a scene. A moment without a character want is simply a moment, an image." and the moment cannot repeat a known emotion; it must show an evolving emotion.
Responsibility for closing the scene："You the writer need to decide how to end the scene. And you the writer need to decide what emotions your characters are feeling."
exposition through conflict: In the HTGAWM forest scene, everyone's opposition is reasonable, and the difficulty and stakes are laid out through the argument.

Five additional specifications at the episode level(the 2-page length target, no need to describe appearance, "ANGLE" as a literary device, multi-layer staging layering within the same scene, don't micromanage actors) andstage directions carrying structural tasksthe two items (Succession script collection) see reference.md Section 8.

7. Season Shape

7.1 The logistical layer and the dramatic layer. Stranger Things S1's logistical three-act 3-3-2 and the minute counts of its dramatic three acts, the season-level pacing discipline of serials, and the episode-count check ("Are you starting the story too early?", cutting 1 episode often speeds things up) see reference.md Section 9.

7.2 Tentpole。 Landau 1e Tip 13 groups teaser / act break / tag together as tentpoles; its functional chain and the role of the tag see reference.md Section 9.

7.3 Bottle episode: the cause is budget. The measured act lengths of West Wing "17 People", the three rules, and "inverting four constraints into three choices" see reference.md Section 9.

A bottle episode does not equal a single storyline. The B/C/D parallelism in Sopranos "Pine Barrens", parallel-editing counterpoint, and the information-decay chain see reference.md Section 9.

7.4 Container episodes and ritual episodes (Succession's two templates). The "limited communication" setup of the pressure-cooker episode, the free beat sheet of the ritual episode, and "the absent protagonist" see reference.md Section 9.

7.5 two-parter。 The measured act pages of West Wing "In the Shadow of Two Gunmen" I/II and the rules for dividing labor between the two halves see reference.md Section 9.

7.6 Season finale and series finale.

Downton's streamers and "the last act only does pay-off", Succession S1's $140/share symmetrical reversal and S4's opposite approach, Sopranos' anticlimactic ending, Sorkin's image list, and the measured act lengths of Two Cathedrals see reference.md Section 9.
Landau 2e: "Give yourself and your protagonist(s) a deadline."——it can also help you finish the pilot.

7.7 Six-episode season shape (Fleabag template). The episode-by-episode skeleton (unfold/desire/exposure/bottle episode/false victory/closed social occasion), the single-episode shape, and two transferable devices see reference.md Section 9.

7.8 Multi-thread season shape (Downton S2). The act counts per episode, the length stratification of storylines, the "retention mechanism", and "before killing a character, first let the audience guess someone else" see reference.md Section 9.

7.9 Alternative structures (Landau 2e). The complete list of Linear / Flashback / bookend / zigzag / hopscotch / three timelines / Rashomon / Time Loop / told backwards, the five questions for choosing a type, and multiple timelines as a tonal regulator (Alex Pina) and Mazzara's opposing view see reference.md Section 10.

8. Divergence retention table: which to use in which situation

Point of contention	Position A	Position B	Criterion
Is the act structure or format	Douglas / Calvisi / Goldberg & Rabkin: four/five/six acts are the structural skeleton, and page-number anchors are verifiable	Oberg: acts are logistical, "just the size of the sausage"; design uses the dramatic three acts, then cut it up at delivery	For platforms with ads, or when you must prove structural competence → A; diagnosing "sagging middle" or "formulaic structure" → B (split the sequence by subgoal). They can be stacked: A as the working unit, B as the design tool
Whether to write ACT on the page	Landau: break the story by act breaks, and delete the labels in the final draft (story over format)	The opposite in the same book: many showrunners want to see the labels as evidence of structural mastery; Miller: if the content can go on a broadcast network, you must write it	Look at the content's target market (Miller's criterion): for a streaming crew you can delete it, for broadcast network/basic cable you must keep it
Whether the midpoint has meaning	Calvisi: the Midpoint is the Act Two curtain; it must reverse the A story and carry a ticking clock	Oberg: "the midpoint can be safely ignored if you find it confusing"; the substitute is mid-act climax	If the A story goal can cover the whole of Act 2 → midpoint; if the goal changes midway (heist films, the Occupied pilot) → mid-act climax
Whether the midpoint is a false victory or a must-succeed	The Snyder school/Calvisi: choose either the illusion of victory or the illusion of defeat	Miller: it must be hero succeeds—almost——"It's not 'the hero fails'!"	When writing a tragedy/antihero episode, use Miller's own variant:squeeze flips positive, climax flips negative
The placement of the "worst-case scenario"	Douglas puts it at the three-quarter point; Calvisi's DOW + All is Lost is at the 4/5 point	Oberg："a series will be successful despite these formulaic crutches, not because of them."	Writing a TV episode for the first time → use the anchor points for self-check; already have solid sequence design → no need to lay them out
A/B/C stories vs other story-breaking methods	Douglas / Calvisi / Landau: A/B/C(/D) + runner, the letters ordered by screen time, and the act break lands on A or B	Mazzara: rejects A/B/C, and instead asks "What is this episode about?"; Wendy West: don't break by act, break by character and then weave	Ensemble drama where each episode should be "a complete story" → Mazzara; single-protagonist serial where the reversals are concentrated in the protagonist's story → West; procedurals and broadcast dramas → A/B/C. The cost is pointed out by Mazzara himself: secondary characters may remain secondary characters forever
Must every act end on a cliffhanger	Calvisi: every act (ideally every scene) should make people want to turn the page; Kelley (Revenge, six acts): "five or six cliffhangers an episode"	Landau 2e: "forcing a cliffhanger at the end of each act break is deadly"; Sorkin's empirical finding: the Act One and Act Two curtains are both very light, and Act Four isn't a cliffhanger but a "reset"	Six-act broadcast drama (five ad slots to fill) → Kelley's quota; no ads or a strong auteur drama → Landau/Sorkin.The criterion doesn't change: the jeopardy/risk/fear must grow out of the character
Unified theme across multiple storylines vs theme as an emergent thing	Landau (Tip 14 / ch11): theme is the glue that holds divergent storylines together (all four storylines of ER's "Stuck on You" are about being "stuck on each other")	Johannessen (Homeland/Dexter/24): 24's room slogan "We don't do themes."；"If you want something that looks and feels real, throw out the writerly devices"; the substitute is operational question. Mazzara is on the same side: focus on the story first, and only step back to see the theme	Strong realism/thriller priority → B (switch to the "operative question of this episode"); ensemble multi-storyline, needing to stitch several unrelated storylines together → A. Note that Johannessen himself admits that Dexter S5's "Atonement" isemergent; Landau also lists shows that do not use a unifying theme (Breaking Bad; Parenthood and FNL "arranged asymmetrically")
Do you need an outline?	Landau 1e："No outline, no paycheck."; Calvisi: "the Story Map is the ultimate outline template."; Miller: first grid, then outline, then write	Mazzara: "I don't use outlines… None of their notes are applicable to script." — the alternative is a two-to-three-page story document plus a beat sheet of two or three words per beat ("Rick kills Shane."); Sorkin: "When I finish one script I have no idea what's going to happen in the next."	Writing episodes on someone else's crew / needing development pay → an outline is mandatory (a contractual milestone); being your own showrunner and able to maintain structure while writing → Mazzara/Sorkin is viable. Mazzara himself drew the boundary: "All script problems… come down to structure problems." See details in `sw-writers-room`
Is a teaser mandatory?	Calvisi / Douglas / Goldberg & Rabkin: the teaser is the first hook	Oberg: only if it serves a dramatic purpose, otherwise "just pointless"; Landau: many shows start directly with Act One	With ads and viewers holding the remote → a teaser is mandatory; strongly serialized shows / binge platforms → can substitute a "short first act + late title card" (Scandal, Modern Family)

9. Diagnostic checklist (ask yourself item by item after finishing an episode)

This episode'score questionCan it be stated in one sentence? (Mazzara: "What is this episode about?")
Does every line have a one-sentencelog line attached to the protagonist (not a guest character) ? Is it a real story, or just a situation?
Does the A story run through every act? Do all act breaks land on the A or B story? Is any act ending on the C story — if so, that line may actually be the A or B story.
Does each act have 4–5 beats and 5–7 scenes? Is the total scene count between 20 (five or six acts) and 28–30 (four acts)? (Scene counts apply only to American one-hour episodes: half-hour comedies have 15–20 scenes [multi-camera] or 40–50 scenes [single-camera], see sw-sitcom-comedy Section 4; domestic 45-minute episodes have 25–35 scenes, see sw-chinese-series-practice. The three numbers differ by two to three times; choose according to "shooting style + runtime," do not scale.)
(This item applies only to American one-hour episodes; skip it for half-hour and domestic dramas) Do the page anchors line up? (Four acts with teaser: 17–18 / 30 / 45 / 60; Calvisi five acts: teaser ≤10, Act Two ends at 30, Act Three ends at 40, Act Four ends at 48; AOP at 37–40; worst case at the 4/5 point) (Same as above: applies only to American one-hour episodes; do not run this item for half-hour and domestic dramas)
Is there a scene over 3 pages that is not a complete beat? A four-page meeting scene is a red flag.
Is any scene there only for exposition? Is any "moment without a character want" being used as a scene?
Does every scene have the five-piece set (want / obstacle / at least three escalations, with a preferred action / decision / resolution)? At the end of the scene, does the audience know where the hero stands emotionally?
Can you clearly say what event each act-out is? Did it grow out of the character's jeopardy/risk/fear? For a "surprise" act-out, has it been set up so that it "should have been visible but wasn't"?
How many of the eight act-out forms did you use?Is at least one scene frozen in place using dialogue-free action or a prop?(And was its grammar taught earlier in the text)?
Does the Act Two break change the nature of the story, or merely raise the danger? Does the end of Act Three turn the protagonist's victory into a new defeat?
Stakes test: for this character, is "doing nothing in the face of the crisis" a viable option? Is the +/− polarity of each act marked? Is there any act that ends on a "win"?
Where is each of the four information tools used? Does the protagonist know more than the audience for an extended period (especially backstory necessary for understanding their motivation)? Is this season's big mystery resolved by the season finale?
Is the teaser a story or a message? If a teaser flashback is used, does it accidentally answer this episode's dramatic question?
Does the final act only pay off and introduce no new information? (Not applicable to Chinese long-form series: the Chinese system requires selling another big hook at the end of every episode, which directly conflicts with this item. When writing Chinese series, this item applies only to the "season finale"; for the end of a single episode, follow the end-of-episode cliffhanger rules of sw-chinese-series-practice .) Before the end of the second-to-last act, is it clear to the audience which accounts remain unsettled? Is there something to "discover" at the end, whilenotpaying everything off at the same time?
Streaming script: after removing the ACT markers, can the invisible acts still be delineated using the triple criteria (story day / location change / shift in the core question)? Does each act ending have a transfer of information or power?
Season shape: is this episode in the right position within the season (six-episode season: setup / desire / reveal / bottle episode / false victory / closed-room settlement)? Is there a symmetrical reversal between the season finale and the pilot? Does the placement of each line simultaneously generate debt for the next season?
Speed-up check: can the coolest thing at the end of Act Four be moved to the end of the teaser? Does this episode have "the scene everyone is talking about the next day"?
Production check: have the ratios of interior/exterior and day/night scenes been calculated? Have you written only DAY / NIGHT? Have you turned production constraints into the three choices of a bottle episode?

10. Workflow: From an episode idea to a delivery draft

Lock down three things before writing（Oberg）：series type（procedural / serial / hybrid / limited / anthology / sitcom）、story-type（plot-led / character-led / theme-led / hybrid / exception）、format(episode length + broad category). If any of these is vague, do not write the outline yet. (For story-type determination, see sw-series-engine-bible。）
Choose a structural comp(Calvisi / Landau, same item): find asuccessfulseries close to you in tone / genre / subject matter / audience and use it as a structural template; if you cannot get the script, watch the show and write a scene list. Clarify four things:Is there a teaser? How many act breaks? How many scenes per act? Is there a tag / epilogue? It is also your comp when pitching.Ready-made sampling table of 47 series(series name | year | format | network / platform | act structure | premise / non-premise / hybrid pilot | showrunner; Landau 1e appendix, 2014 standards) see reference.md Section 11; there is no need to recount it yourself.
Break down two or three episodes of the comp and fill in the grid(Douglas): name them A/B/C, write one log line per line anchored to the protagonist, and fill in the letters cell by cell.
Define this episode's core question, as well as each line'sentry scene and exit scene(Downton's line-table method: 8–12 lines is the upper limit for one episode, and 1–2 of them must be unrelated to the main crisis).
Reverse-engineer(Douglas): First fill in the boxesthe opening, the "worst case" at the three-quarter point, the endingthree places; do only the A line or A+B, leave room for C; then work backward from the cliffhanger to the preceding beats.
Do a landmark or dramatic sequence for each line separately: use Miller's six landmarks (the pilot's hero committed must kill two birds with one stone), or Oberg's inciting incident → subgoal → sequence → scene.At this point write each line separately, do not braid.
Compress into a one-page grid(Miller Chapter 5): a top row of act label cards, under each act a column of cards, one card per scene, with the scene heading + character want. For a no-ad show, first use teaser + four acts breaks, delete the markers in the final draft; if you want six acts, split acts three and four each in half. Check at a glance: are the landmarks in place, are the act breaks hard enough, is the protagonist proactive, does this episode fit the season.
Arrange the act outs: hand each act out to a different line; choose from the eight types, make sure at least one act ends with no dialogue or with a prop; keep the squeeze in the middle of the act (it needs to be played out).
Give each episode a "procedure line" as a metronome(the common feature of the six West Wing episodes): an external crisis (military strike, embassy siege, surgery duration, vote count) provides legitimate interruptions and an "one hour left" deadline, and in the end its number is handed to the protagonist as a moral argument; it does not carry an act out.
Write an outline / beat sheet: for one-hour drama use a treatment or step outline;Blum gives series treatments of 5–10 pages, "the shorter the better," and must write out the act breaks(a film treatment, by contrast, is 10–20 pages); write each "beat" of the outline as the log line for that scene. For the room process see sw-writers-room。
Braid only when writing the teleplay: do multi-layer staging within the same scene, ellipsis, parallel editing; at this point add the five-piece set to each scene.
Write long then compress, or stick to the page count: Succession's approach is 70+ pages per episode (only then is there room to refocus in editing); Calvisi's counter-warning is that newcomers should not exceed 60 pages. Which one to take depends on whether you are the showrunner and whether the show has already started shooting.
Make one decision about format before delivery: mark ACT or delete ACT (according to the criteria in 2.6); write the teaser act out as SMASH CUT TO: MAIN TITLES, and at the end of each act FADE OUT. ＋ END OF ACT N(the six West Wing episodes are 100% consistent);the pilot script does not carry scene numbers（Miller："Pilot scripts do NOT have scene numbers."）。
Run the diagnostic checklist in section nine, then use Grace Paley's ending method (as recounted by Landau)—"I look for the lies."

Eleven, Training

Break down an episode and fill in the grid(Douglas's classroom assignment): name the A/B/C lines, write a one-sentence log line for each, fill the letters into the grid cell by cell; mark which line each act-end cliffhanger belongs to, which line only starts in which act, and which one ends by the third act.
Goldberg & Rabkin's double exercise(p0033): ① Watch an episode, stop at each act,write the narrative arc of that act in one sentence, then write the last scene of that act in one or two lines; ② Watch the Act Two breakof several episodes in a row, and analyze how that moment changes the direction of the story.
act out count: take one episode, write out which of the five act outs belongs to which type (the eight-type table), and judge which one is schmuck bait.
Calvisi's Story Map: use the Basic 17 columns + Full map to break down a pilot, mark which page the AOP falls on and whether the DOW is at 4/5; then beat out your own pilot using the same chart.
Miller's landmark form: write six lines for your own pilot; then break down astructure(not subject matter) similar to your pilot; the breakdown must include scene headings, story beats, character wants, scene ins and outs, act breaks, landmarks.
escalation breakdown: pick two 1.5–2 minute scenes, count out the escalation item by item (aim for 10+), mark the decision and resolution, and tally how much is action versus how much is dialogue.
Reverse-engineering invisible acts: take an episode of a show with no ACT markers (Succession, Homeland, Sopranos), use triple criteria to mark the act positions, write out "act-out event + basis for the division" for each act, and compare with reference.md section 4's Succession hands-on comparison.
Information management rewrite: rewrite a "cut away mid-fight" act out into three versions—one surprise, one newly established dramatic irony, one mystery; judge which makes people want to watch the next episode more.
Act count conversion: convert a four-act episode outline into five acts (split Act Four in half), six acts (split Acts Three and Four each in half), and no acts (change to story-day markers); each time you must find a defensible event for the new act break.
Speed-up exercise(Rabkin): move the coolest thing at the end of Act Four of the pilot to the end of the teaser, rewrite the whole episode, and see whether the next sixty pages can maintain the same pace.
Budget-driven: impose four prohibitions on yourself (no guest stars, no exteriors, no new sets, no crowd), write a bottle episode—single night, single building—use it to advance the secret in the season arc you haven't dared to reveal, and increase the line count from 4 to 5.
Six-episode season shape: write one-sentence outlines for six episodes following the Fleabag skeleton, designate one object thatchanges hands each episode; the final image of the season finale resolves only the theme.
Symmetry exercise(Succession S1): find a number/object/line for the first episode, and have the season finale use the same thing to complete a reversal.
Container episode exercise: write an episode where "everyone is locked in one building + communication is limited," then write an episode that is "a ritual day with a schedule" (list the established segments, and insert one conflict into each segment).