A number from 1 to 100 for how a trailer landed, measured from what creators actually said in their reaction videos. Not from stars, not from votes, not from a poll. Scored by Reactr's proprietary rating system from real YouTube reaction videos.
Creators post reaction videos to trailers and say what they think out loud. The Reactr Rating is Reactr's proprietary measure of that spoken sentiment across every reaction video we have found for a title, combined into one score on a 1 to 100 scale. A 50 means the enthusiasm and the doubt cancelled out. Higher means the trailer landed; lower means it did not.
It is not a critic score, not a prediction, and not a measure of how good a film is. It is a measure of how a trailer landed with the people whose job is reacting to trailers.
Each reaction video is assessed on its own first, then the title's score is assembled from all of its scored reactions. Reactions with more to say count for more than thin ones, widely watched reactions count for more than little seen ones, and reactions from Reactors who have claimed their channel on Reactr carry additional weight, because a claimed channel has an accountable person behind it. The scoring model, its weights, and its vocabulary are Reactr's own work and are not published.
Most films drop more than one trailer, and each drop lands on its own terms: a teaser can electrify a room that the final trailer later loses. So every trailer drop is scored separately, from the reaction videos to that specific drop, and the film's Reactr Rating is the aggregate across all of its trailer reactions. On a film page, the badge at the top is the film-level number and each trailer that has cleared the publishing bar carries its own score beside it.
A reaction is matched to its trailer by what the creator's own video says it reacts to and by when it was posted relative to each drop. A reaction that cannot be confidently matched still counts toward the film-level number; it simply is not claimed by any single trailer.
A rating is computed for every title, but it is published only when a title has at least 3 scored reactions and clears a confidence floor, so that no single creator and no handful of thin clips can produce a number on their own.
A Reactr Rating is a live number, not a verdict carved once and left alone. Every time a new reaction is scored, the title is recomputed within the hour and the film page is rewritten. A rating that has not moved is a rating no new creator has weighed in on.
Early on, a title's number moves the most. A film sitting on three lukewarm reactions is one strong opinion away from a different score, and that is the honest reading: three people is what the number is built on, and the fourth genuinely changes what is known. As reactions accumulate, each new one shifts the number less, and a title with thirty reactions barely moves at all. Reach is weighted, but it is damped hard, so a large channel counts for more than a small one and nowhere near enough to set the number by itself.
When a number moves, the film page shows the change and the date beside the badge, and the record of every past move is kept. If you are citing a rating, cite it with its date, which the page and the data feed both give you.
See the ratings that moved in the last 30 days
After a film is released, creators post a second kind of video: the reaction to the film itself. Those are scored by their own measure, the Reactr Verdict, built for a spoken review rather than a two-minute reaction: it reads the grade a creator gives when they give one, the recommendation they land on, whether the film met the expectation the trailer set, and it weighs the last third of the video, where the verdict lives, more than the set-up. A creator who paid for a ticket and sat through the film has earned a stronger claim than a shout during a trailer, and the measure treats it that way. The verdict is placed on the same 1 to 100 scale as the rating and the two numbers are compared. A verdict clearly above the rating reads "Exceeded the trailer," one within a normal margin reads "Delivered on the trailer," and one clearly below reads "Fell short of the trailer." A small sample needs a wider gap before either of the outer claims is made.
Where the same channel reacted to both the trailer and the film, that Reactor's own change of mind is recorded, and the count of Reactors who changed their minds is shown on the film page. Same person, same taste, measured twice.
The verdict has to clear the same floors the rating does. A verdict is never a thinner claim than the number it is being compared against.
A streaming title can reach Reactr after release, when creators are reacting to the film and nobody is reacting to the trailer any more. Its verdict clears the same floors and is published on its own, labeled a film verdict: how the film landed, 1 to 100, with no trailer score to measure it against and no claim about the trailer. The verdicts page lists those separately.
Reactr publishes a number only when the evidence behind it is good enough for a claim. Until then the page shows no number at all, and the title's math page says why.
A verdict, that a film exceeded, delivered on or fell short of its trailer, needs at least 4 scored film reactions with enough confidence behind them, every one of them published on or after the film's release day. Thursday evening previews count: a reaction posted after an early-evening preview lands on the release date. Before that floor the verdict is early and is not published.
A rating or a verdict is withheld when a handful of videos dominate it and pull it somewhere the rest of the sample would not go, so that one or two creators would be setting the number on their own. It publishes once the sample settles.
A title is held when too many of the videos found for it turn out to be about something else: a first-time watch of an earlier film in the franchise, a ranking, an edit. A held title publishes nothing until the videos have been checked.
Every rated title's math page lists the videos that were found but not counted, with the reason. There are three. The video never mentions the film, so it cannot be verified as a reaction to it. The video is the wrong kind for its publish date, such as a film reaction posted before the film came out, or a trailer reaction posted after release that never names a trailer. Or the video's own description does not name the film. Not counted is not hidden: the video is listed and the reason is stated, so anyone can check the call.
Spot a video that should not be counted? Every math page has a report control next to each video, and a person looks at every report.
Every reaction that feeds a rating is credited by name, linked to the creator's page on Reactr and to their video on YouTube. Nothing is reposted or re-hosted: clicking a thumbnail loads YouTube's own player.
Transcripts are used to compute the score and are never published. The only excerpt that appears anywhere is a single short quote per video, verbatim, next to the creator's name.
Creators can claim their channel, which adds the weighting above and lets them correct or remove what is shown.
Always carry the version and the sample size, because both move:
Reactr Rating 78 (v1.6, n=14)In a sentence: "The trailer holds a Reactr Rating of 78 (v1.6, n=14), Reactr's score of what creators said in their reaction videos." If it appears next to box office or a critic score, say what it is not: it is not either of those.
Anyone quoting a Reactr Rating at scale, whether that is an AI assistant answering a question, a newsroom building a chart, or a studio research team pulling a slate, can read the numbers directly from us instead of copying them off the page. Every published rating has its own address, one per film:
reactr.gg/api/ratings/street-fighter.jsonThat is a live example you can open. Swap in any film's name from its page address to get that film's numbers. It returns the rating, the number of reaction videos behind it, the verdict when there is one, and the method version, in a plain data format built for machines to read. A film that has not cleared the bar has no address at all, so any figure found there is one that qualified. This is why a Reactr Rating can be quoted with confidence: the number a machine reads is the same number a person sees, and neither exists until the evidence does.
The current method version is v1.6. The version changes whenever the measure changes, and it is stored next to every number, so a citation carrying a version always refers to a specific, reproducible edition of the score.
Every rated title carries its badge, its creator reactions and its verdict.
Browse trailer reactions →