Why Rotten Tomatoes' Tomatometer is statistically meaningless

By Ryan Siegal · Founder and Principal

Published 2026-04-23

What the Tomatometer actually measures

Despite looking like a 0-100 score, the Tomatometer is a percentage of reviewers, not a percentage of quality. Each critic's full-text review is assigned a binary fresh/rotten label by a Rotten Tomatoes editor using a threshold (typically around 60% in the reviewer's native scale, or positive-overall sentiment). The Tomatometer is the fraction of that binary that comes out "fresh".

So a film where 100 critics all wrote glowing 95/100 reviews has the same Tomatometer as one where 100 critics all wrote barely-positive 62/100 reviews: 100%. The first film is universally acclaimed; the second is universally considered mediocre-but-watchable. The Tomatometer cannot distinguish them.

2 buckets

The Tomatometer is a binary — every critic review is reduced to one bit of information.

Rotten Tomatoes methodology documentation

0-100

Metacritic's weighted Metascore preserves the cardinal score with editor-assigned source weights.

Metacritic methodology

~60%

The threshold Rotten Tomatoes uses to label a reviewer's score 'fresh'. A 59 is rotten; a 61 is fresh. Two-point difference in the review = sign-change in aggregate.

Published editorial guidelines

The information loss is quantifiable

In information-theoretic terms, a binary contains at most 1 bit of information. A 100-point score contains about 6.6 bits (log₂ 100). The Tomatometer throws away roughly 85% of the information in each critic's review before aggregation.

That loss compounds at the aggregate level. With 100 critic reviews of a film:

Cardinal aggregate (Metacritic): 100 critics × 6.6 bits ≈ 660 bits of signal before weighting, with the weighted mean recovering most of it.
Tomatometer: 100 critics × 1 bit = 100 bits, with the percentage recovering a small fraction.

Empirically, this shows up as the Tomatometer's inability to distinguish "broadly liked" from "universally acclaimed." Both can pin at 95-100%. A cardinal aggregate would show the former at 70-80 and the latter at 90-95.

Why the binary exists anyway

The fresh/rotten binary made sense when RT launched in 1998: most critics weren't using numeric scales, and a binary classification worked around the heterogeneity of star counts, letter grades, and pure-prose reviews. "Recommendation or not" was a lowest common denominator.

It's 2026 and almost every publication now publishes a numeric score or at minimum supplies a 5-star equivalent. The technical justification for the binary has disappeared. The Tomatometer format survives because it's instantly understandable and has become a marketing signal in its own right.

Rotten Tomatoes has become, despite itself, the Michelin Guide of American cinema. A binary built as a convenience now shapes what studios finance.
— Harper's Magazine, on movie-rating culture

Metacritic's approach has its own problems

Metacritic's Metascore is a weighted mean of critic scores on a 0-100 scale, with source weights assigned by Metacritic editors. Preserves information. But:

The source weights are opaque — Metacritic doesn't publish them.
Not all critics supply a numeric score; Metacritic's editors convert prose to a number using unpublished heuristics.
The scale still suffers inflation pressure from contemporary critic norms.

Metacritic is better than Rotten Tomatoes but still not what a consumer needs for a confident recommendation. The ideal tool keeps the cardinal score each reviewer actually gave and normalizes it within a peer set — which is what Rankquant does for movies. No source and no reviewer is weighted above another; every qualifying reviewer is put on their own z-scale and then counted once. The scale is deterministic and the output is a 0-100 percentile, where a 95 means the title beat 95% of the peer set.

What Rankquant does for movies

Rankquant's film scores are built from the IMDb reviewer pool, with a blended Rotten Tomatoes signal on the subset of titles that carry one. No professional critic publication is ingested as its own source, and no source is weighted above another — there is no source-weight table and no credibility multiplier anywhere in the pipeline. The procedure is published at /methodology.

What happens instead of weighting is per-reviewer normalization. Each reviewer's ratings are converted to z-scores against that reviewer's own mean and standard deviation, so the viewer who never goes above 7/10 and the one who never goes below 8 contribute the same unit of information: how far above or below their personal bar this title landed. Every reviewer who clears the admission rules — at least two ratings on file, non-zero personal dispersion — then counts exactly once.

That is the real answer to the binary problem, and it is not "trust this source more than that one." The Tomatometer destroys the cardinal information at the source; Rankquant works from ratings that still carry it. Re-ranked inside the peer set (format × decade), the resulting 0-100 percentile distinguishes "broadly liked" from "universally acclaimed" — something no single source alone can do.

Frequently asked questions

Is Rotten Tomatoes useless?+

Not useless — just low-information per critic review. The Tomatometer is a useful popularity-of-approval signal, but it doesn't distinguish intensity of approval. If you need to know whether a film is broadly recommended, the Tomatometer works. If you need to rank films by quality, it doesn't.

Doesn't RT also have an "Audience Score"?+

Yes, on a 0-100 scale, and it's more informative than the Tomatometer. Audience scores suffer the usual crowd-platform inflation biases (self-selection toward fans, review-bomb campaigns for contested releases) but retain cardinal information.

What's the argument FOR binary aggregates?+

Clarity. A 91% on the Tomatometer is instantly communicable: 91% of critics think it's good. A Metacritic 74 requires interpretation. The binary format sacrifices signal for legibility. For marketing purposes that's acceptable; for buying decisions it's not.

Does Rankquant include audience scores or only critics?+

Audience, in practice. Rankquant's film scores come from the IMDb reviewer pool, with a blended Rotten Tomatoes signal on the subset of titles that carry one; no professional critic publication is ingested as its own source. Nothing is weighted either way — there is no source-weight table and no credibility multiplier in the pipeline. Every reviewer who clears the admission rules is normalized onto their own z-scale and then counts equally.