Filed August 31, 2026. Resolution date September 6, 2026.
1. Why this paper is dated
Every claim in this paper about the future is made before the data exist. That is the paper’s only real contribution, and it is worth stating at the outset why so modest a thing should count as one.
The standing literature on word of mouth in entertainment is almost entirely retrospective. A record sells poorly and a reference work explains that it suffered from bad word of mouth. A film holds well and the trade press credits strong word of mouth. In neither case was the mechanism observed. It was inferred from the outcome it was then used to explain, which makes it unfalsifiable in the plain sense that no result could have contradicted it. Had the record sold well, the same reference work would have credited good word of mouth, and no sentence would have had to change except its adjective.
An earlier paper in this series took up that problem in the case of Mike + The Mechanics’ Word of Mouth (1991), where an attribution of poor informal reception was made after the fact and without evidence that any such reception occurred. The difficulty there was structural and could not be repaired by better research, because the relevant conversations were never recorded and the album’s commercial history is the only trace anyone has. The present case repairs it by accident. Here, for once, the prospective instruments exist, they have already reported, the outcome will be public within a week, and the question of what word of mouth is doing can be asked in a form that admits an answer of no.
So this paper states in advance what number would confirm the claim that word of mouth is carrying this film, what number would refute it, and what number would leave the question open. If the prediction is wrong, it will be wrong in public and on the record. That exposure is the point. It is also the same standard the earlier paper proposed for informal speech generally, applied here to scholarship rather than to gossip: a claim is worth something in proportion to what its maker stands to lose by making it.
2. The case
Coyote vs. Acme was completed at a reported production cost of around $70 million and then shelved by Warner Bros. as a write-off. After public reaction to the prospect that the film would never be released, Ketchup Entertainment acquired the rights for $50 million. Marketing spend has been reported at roughly $10 million, thrifty by the standards of a wide release, and the film opened on 3,575 North American screens. Foreign distribution rights were sold for a reported $20 million, which places domestic breakeven for Ketchup somewhere near $40 million.It opened to an estimated $15.5 million domestically, second behind the fifth weekend of Spider-Man: Brand New Day at $22.2 million, with $8.4 million internationally for a $23.9 million global start.
Friday accounted for $5.3 million of the domestic figure, inclusive of $1.7 million in Thursday previews. The prospective reception instruments were unusually strong: an A CinemaScore, 93 percent positive and 82 percent definite recommend on PostTrak, against 95 percent approval from critics on Rotten Tomatoes.
Two comparisons frame the number. It is by a wide margin the best opening in Ketchup’s history, against a previous high of $3.1 million for The Day the Earth Blew Up, another Warner Looney Tunes castoff. It is also softer than the recent norm for live-action and animation hybrids, the last fifteen of which averaged around $23.2 million. One box office analyst characterized the numbers as fair and the reception as excellent, expecting recovery through ancillary rather than theatrical.
That is a film whose theatrical fate is genuinely undetermined and whose commercial case rests, by everyone’s account including its distributor’s, on how it holds rather than on how it opened.
3. Two populations with opposite time signatures
The interesting feature of this case is not that word of mouth matters to it. Word of mouth matters to most films. The interesting feature is that this film’s opening audience is known to contain a population whose attendance is not a report on the film at all.
A visible constituency wanted Coyote vs. Acme to succeed as a rebuke to the executive who shelved it. Its members bought tickets in order to affect a number that would be reported as a verdict on his judgment. Their enthusiasm is real, their money is real, and their speech about the film is not evidence about the film, because it was determined in advance of the screening. Had the film been poor, that speech would not have reversed; the failure would have been reassigned to the marketing budget or the release date. A subset of this constituency also profits from the controversy in visibility and income, and their interest runs specifically toward the fight continuing rather than toward the film being good, since a quiet well-liked picture that earns its money is worse for them than a running dispute.
The methodological consequence is that the opening weekend figure is contaminated in a way that cannot be corrected by any adjustment to the figure itself. But it can be separated in time, and this is the paper’s central claim:
A vindication audience is front-loaded by construction. Word of mouth is back-loaded by construction. The second weekend is therefore the instrument that distinguishes them.
The reasoning is straightforward. People attending to register a verdict attend immediately, because the opening number is the reported number and the one that carries the argument. They have no second visit to make and no one to recruit who is not already recruited, since their network was mobilized before release. Word of mouth in the ordinary sense requires a person to see the film and then persuade someone who had not planned to go, which takes days and works outward. The two mechanisms therefore have opposite curves, and a film driven by the first decays quickly while a film driven by the second decays slowly or grows.
This is what makes the prediction non-trivial rather than a restatement of the CinemaScore. A film can have an A from the people who saw it and still collapse, if the people who saw it were the only people who were ever going to see it. The A tells you the film satisfies the population that shows up for it. The hold tells you whether that population extends beyond the mobilized one.
4. The prediction
Weekend two runs Friday September 4 through Sunday September 6, with Labor Day falling Monday September 7. Both the three-day and four-day figures should be recorded, but the three-day is the test figure, since it is comparable to the opening frame.
Baseline expectations. A wide family title with an A CinemaScore normally declines somewhere between the high twenties and low forties in its second weekend. A Labor Day frame conventionally improves the three-day comparison by something on the order of ten percentage points, because Sunday plays as a second Saturday when Monday is a holiday. Back-to-school scheduling in most of the country pushes the other way for a family title in this specific week. These figures are rules of thumb and are stated here as the paper’s assumptions rather than as findings; a comparison set of hybrid and animated family titles holding into Labor Day should be assembled and reported alongside the resolution.
Thresholds, against a $15.4 million three-day opening.
Strong confirmation. A decline of 22 percent or less, meaning $12.0 million or better. This would mean the film reached an audience in week two that was not present in week one, which is what the word-of-mouth claim asserts and which the holiday alone does not deliver.
Consistent with the claim. A decline of 22 to 32 percent, $10.5 to $12.0 million. Good, but partly attributable to the frame. The claim survives without being demonstrated.
Neutral. A decline of 32 to 42 percent, $9.0 to $10.5 million. This is the ordinary result for a well-liked family picture in a holiday weekend and licenses nothing. Any word-of-mouth story told about this range is the retrospective attribution this paper exists to avoid.
Refutation. A decline past 45 percent, under $8.5 million. This would indicate that the opening was a demonstration rather than a market, that the mobilized constituency exhausted itself in the first frame, and that the general audience behind it was thin. In a holiday frame, a drop of this size for an A-rated family title is a strong signal and should be reported as a refutation without hedging.
Secondary instruments.
The weekday grosses for Monday August 31 through Thursday September 3 arrive before resolution and are the earliest available tell. A front-loaded profile shows steep weekday decay; a film building word of mouth shows unusually flat weekdays for its size.
The final domestic multiplier, total domestic divided by the $15.5 million opening, is the more meaningful number and should be predicted now even though it resolves in October. If the 82 percent definite recommend means what the industry treats it as meaning, the implied multiplier is above three, giving a domestic finish above roughly $47 million. A multiplier at or under 2.3, roughly $35 million, is the front-loaded outcome. Note that this threshold nearly coincides with the reported $40 million domestic breakeven, so the word-of-mouth question and the solvency question have close to the same answer, which is a coincidence worth flagging as a hazard: it will be tempting after the fact to let the commercial verdict stand in for the empirical one.
5. Pre-registered confounders
Naming the excuses in advance is what keeps this from becoming another retroactive attribution with better manners. The following are specified now and may not be added later.
Competition. September 4 brings several new titles, including Tom and Jerry: Forbidden Compass and a Cars twentieth anniversary re-release, both of which compete for the same family audience. Their theater counts and results must be reported with the resolution. If a competing family title opens wide and takes screens from Coyote vs. Acme, the drop figure is confounded and the paper should say so rather than either ignoring it or hiding behind it.
Theater count. A hold at 3,575, a contraction, and an expansion are three different tests. The count must be recorded and the per-theater average reported alongside the gross.
Back-to-school. Working against the film and partially offsetting the holiday. The two effects cannot be cleanly separated with the data available, which is a limitation of the test rather than a reason to discount either.
Distribution announcements. A premium home release date announced before or during weekend two voids the run test, since it changes the audience’s reason to wait rather than go.
Estimates versus actuals. The Sunday figures are studio estimates. Monday actuals sometimes move a point or two. Resolution should use actuals.
6. What a good hold would and would not show
If the film holds at 20 percent, the correct conclusion is narrower than the one the trade press will draw.
It would show that the prospective instruments predicted the outcome. CinemaScore and definite recommend are measurements of stated intention to recommend, taken at the theater on opening weekend, and the hold is the behavior those measurements are supposed to anticipate. Establishing that they did anticipate it in a case where the opening was contaminated by a non-representative constituency is a real finding about measurement, and it is the finding this paper is entitled to.
It would not show that word of mouth caused the hold. The counterfactual is unavailable. A film with these reviews and this novelty might have held on the strength of its own reputation, its critical reception, the holiday, and the absence of comparable competition, with no interpersonal recommendation required. The claim that people told other people to go, and that those people went, remains untested by any box office figure whatsoever.
This limitation is worth stating loudly, because it is precisely the step that retrospective accounts skip. The industry claim is causal. The evidence available is predictive at best. Admitting the gap is more useful than closing it by assertion, and it is the same discipline the earlier paper asked of institutions that attribute their fortunes to informal speech.
There is a further asymmetry in what the test can deliver. Refutation is cleaner than confirmation. A steep drop rules the word-of-mouth story out, because no interpersonal recommendation mechanism produces a collapse. A shallow drop merely fails to rule it out among several live alternatives. The test is therefore better at destroying the claim than at establishing it, which is the ordinary condition of empirical work and not a defect.
7. The fused act
The ethical strand of this case is separable from the empirical one and does not depend on how the prediction resolves.
An institution that prizes word of mouth as free advertising while treating word of mouth about its internal conduct as gossip depends on being able to keep the two apart. Ordinarily it can. Praise for the product arrives through one channel and complaint about the company through another, and the institution can welcome the first as generosity and receive the second as grievance, without ever having to defend the distinction, because the categories rarely touch.
Here they are fused. The film exists in theaters because people spoke publicly about a decision the company considered internal, a write-off of a finished asset. That speech is exactly the kind an institution classifies as leak, rumor, or outside agitation about a matter properly its own. It created the market that Ketchup then bought into. And for a portion of the current audience, the purchase of a ticket is the utterance about the company’s management. The commercial act and the accusatory act are one act, performed once, by the same person, at the same moment.
This is the condition under which the convenient distinction fails outright. Warner cannot accept the box office as a neutral market fact and dismiss the accompanying speech as gossip, because the box office is the speech. Nor can the distinction be repaired by pointing out that the accusers are not disinterested, since the same is true of every favorable word of mouth an institution has ever welcomed. Interest disqualifies nothing on its own; if it did, no praise would count either.
What the case does show is that interest and cost come apart. The earlier paper proposed cost as a proxy for credibility, on the theory that a speaker who pays and gains nothing has no reason to misreport. The vindication constituency defeats that proxy. Its members pay, in tickets and time, and they gain anyway, in a verdict they were already prosecuting. Nothing is disclosable, since nobody paid them, but the speech is not responsive to the film. The film is the instrument.
The better test is the counterfactual one, and it has the advantage of being checkable from outside: would this speaker have said the opposite if the opposite were true? A reviewer with a record of unfavorable judgments about things he is sympathetic to has answered that question in public over time. A constituency organized around a predetermined verdict has answered it too. Cost is invisible and easy to claim. A record is not.
Finally, the gratitude problem. The filmmakers have thanked their online advocates for the visibility they generated, which is a decent gesture and also an accurate description of unpaid marketing labor. But the gratitude cannot distinguish among the three populations it addresses: the constituency arguing about a studio executive, the commentators whose income depends on that argument continuing, and the general audience that has no view on any of it and whose behavior alone determines the second weekend. The distributor has no way to solicit the third kind of speech without also rewarding the first two, and no standing to complain about the first two given what they delivered. That is not hypocrisy. It is the position an institution is in once it has accepted a benefit produced by speech it would have preferred to suppress.
7a. The Financed Advocate
Added August 31, 2026, prior to resolution. The thresholds in section 4 are unchanged.
The preceding section described a fusion of the commercial act and the accusatory act: for part of this film’s audience, buying a ticket was itself the statement about Warner’s management. That fusion arose from circumstance. It was nobody’s design, and no institution could have engineered it.
An institution did engineer something adjacent to it, and the paper must account for that before resolution rather than after.
The structure
In May 2026, Ketchup Entertainment announced a partnership with Legion M for this film’s release. Legion M is an equity-crowdfunded entertainment company operating under the JOBS Act, launched in 2016, with more than 60,000 shareholders, followers in the hundreds of thousands, and a stated ambition of uniting a million fans as owners. It has participated in the release of more than fifteen films in partnership with distributors including Neon, Searchlight, and Bleecker Street. Its function in those partnerships is described by its own account and by trade coverage in the same terms: it augments the distributor’s marketing through direct contact with ticket buyers, channeling the preferences and energy of fans into a difficult theatrical market.
Two mechanisms operated here. The first is the standing one. Every Legion M shareholder holds a stake in the company, and the company’s project page states without qualification that every investor has a stake in this movie’s success. The second is specific. The Legion M Film Fund offered direct participation in this film’s prints and advertising, and that offering sold out in under 72 hours. The distributor’s marketing spend, reported at roughly $10 million and characterized throughout the trade coverage as thrifty, was therefore financed in part by members of the audience it was designed to reach.
Both parties described the arrangement candidly and in advance. Ketchup’s chief executive said the film had already inspired a groundswell of passion from fans and that a community-driven approach was the right way to champion its release. Legion M’s founders said the film was the kind of project their company exists for, and that the fanbase which had fought to rescue it could now own part of it. The trade press reported the arrangement as what it is: a mechanism for converting existing enthusiasm into ticket sales.
Nothing here is concealed. The structure is the pitch.
The fourth population
Section 3 distinguished three populations: the constituency prosecuting a case against a studio executive, the commentators whose income depends on that case continuing, and the general audience that has no view on any of it. The fan-investor belongs to none of them.
He is not primarily arguing about Warner’s management, though he may sympathize. He is not monetizing the controversy, and would generally prefer it resolved in the film’s favor and forgotten. And he is not disinterested. His interest is financial, direct, modest in scale, and published.
He is the harder case, and the paper is better for having to answer him, because he defeats every criterion this series has proposed except one.
Cost fails. The cost criterion held that a speaker who pays and gains nothing has no reason to misreport. The fan-investor pays twice, once for equity and once for the ticket, and gains regardless. Cost turns out to measure sincerity of attachment, not independence of judgment, and those are different properties. A person may be entirely sincere and entirely uninformative at once.
Disclosure fails. This is the finding worth carrying forward. Disclosure is the standard remedy for interest, and the assumption behind it is that concealment is what makes interested speech dangerous. Legion M conceals nothing. Its investor count is public, its partnership was announced in the trades, its film fund was advertised, and its shareholders describe themselves as shareholders in the comment sections of the very articles announcing the deal. And the speech the arrangement produces is still not evidence about the film, for exactly the reason the vindication constituency’s speech is not: the position was fixed before the screening. Disclosure tells a reader what to discount. It does not restore the information that was never there.
Absence of gain fails outright and requires no argument.
The counterfactual test survives. Would this speaker have said the opposite if the opposite were true? For a shareholder writing about a film he owns a fraction of, that question cannot be answered from the utterance, from his balance sheet, or from his good faith. It can be answered only from his record: a history of unfavorable judgments about things he is invested in, sympathetic to, or fond of. Such a record is public, accumulates slowly, is expensive to fake, and cannot be produced on demand at the moment it is needed. Most members of the constituency do not have one. This is the criterion the series should carry forward, and section 7a is the reason.
What it does to the prediction
Section 3 argued that a vindication audience is front-loaded by construction while word of mouth is back-loaded, and that the second weekend separates them. That argument was made about a diffuse cultural phenomenon. It now has a named and financed mechanism.
Legion M’s declared function in this release was to convert enthusiasm into ticket sales through direct contact with buyers, executed across a shareholder list of sixty thousand and a follower base an order of magnitude larger, aimed at the opening frame because the opening frame is where such campaigns are aimed. Whatever fraction of the $15.4 million that represents, it is front-loading of the most deliberate kind: organized, funded, professionally run, and concentrated by design in the days the paper identified as contaminated.
The paper therefore records, before resolution, that the refutation band in section 4 is more likely than it appeared when the bands were drawn. The author’s expectation has shifted toward a steeper decline.
The bands do not move. They were set on August 31 before this factor was recognized, and they stand as set. A pre-registration that revises its own thresholds upon learning something that bears on the outcome is not a pre-registration; it is a retrospective attribution with a date stamp, which is the practice this entire paper exists to refuse. Recording a changed expectation while leaving the test fixed is the whole discipline, and it is worth noting that the temptation runs the author’s way: as a Legion M shareholder, disclosed in section 8, he has an interest in the confirmation band and is on record declining to widen it.
What this does not show
The claim here is narrow and should not be inflated in either direction.
It does not show that the film’s reception is manufactured. Legion M activates people who already want to see a film; it does not produce enthusiasm from nothing, and the enthusiasm around this title predated the partnership by nearly three years.
It does not compromise the CinemaScore or the PostTrak figures. Exit polling samples people leaving the theater, the overwhelming majority of whom have never heard of Legion M, hold no shares in anything, and came because a Looney Tunes movie was playing. That population remains the one the second weekend depends on, and its verdict remains the cleanest evidence about the film available in this case.
It does not make the arrangement improper. Fan equity is a legitimate financing instrument, disclosed and regulated, and the film reaching theaters at all is partly its doing.
What it shows is narrower and stranger. A distributor with a thin marketing budget partnered with a company whose product is the enthusiasm of its shareholders, and a measurable share of this film’s word of mouth was therefore issued by people holding a documented financial stake in its box office. Every one of them would say so if asked. Several said so unprompted, in public, in comment threads beneath the announcement.
That is the condition the earlier papers in this series kept circling without naming: informal speech has become an asset class. Not corrupted, not bought in the sense that requires concealment, but capitalized, with the speaker’s interest converted into an ownership share and the ownership share offered as a reason to speak. The categories the reference works rely on, gossip against testimony, promotion against report, insider against audience, do not survive it. Neither does disclosure, which was the remedy of last resort.
What survives is the record. Whether a person has ever, in public, said the unwelcome thing about something he loved.
8. Disclosure
The author saw the film in a theater near Sandy, Oregon on August 30, 2026, paid for the ticket, liked it, and published a favorable review before writing this paper. He is a small investor in Legion M, who partnered with Ketchup! to release and market the film and did not participate in the direct investment into the film itself, and no position on the management of Warner Bros. He is aware that a person who has praised a film and then designs a test of whether praise like his is doing any work has an interest in the affirmative outcome, which is why the thresholds above were set before the weekday numbers were available and are not to be moved.
9. Resolution
Resolution is due Monday September 7, 2026, on actuals. The resolution note should report the three-day and four-day grosses, the percentage decline, the theater count and per-theater average, the weekday figures for August 31 through September 3, the competitive set with theater counts, and a plain statement of which of the four bands in section 4 the result fell into. If the result refutes the prediction, the refutation goes at the top of the note.
Companion pieces in this series address the chart-data test for retrospective attribution, reference works as institutions, category collapse as an enforcement technology, and the unit of judgment in informal speech.
