The Best Anime Studios by Codex Average: Wit Edges Madhouse on a Three-Show Sample, and Reputation Doesn't Survive the Spreadsheet

Ranking the best anime studios by mean Codex score across their full ranked catalogue rewards consistency over volume — and the order it produces is not the order the reputation economy assumes.

Kyoto Animation isn't in the top eight. Ufotable isn't either. The studios that win on a flat average of the Codex rubric are the ones that either kept their catalogue small enough to curate or kept their batting average high across decades of output — and exactly one studio on the leaderboard has done both at scale.

Reputation and output volume get conflated. A studio with forty shows and three masterpieces gets called a titan; a studio with three shows and three strong ones gets called a boutique. Averaging one rubric — story, character, themes, world-building, animation, cultural impact — across every ranked show a studio has made is a fairer test of whether the reputation holds. MyAnimeList aggregates user sentiment per show and stops there. Fan discourse rewards whichever studio is animating the loudest seasonal title. The Codex catalogue, scored by one consistent six-axis framework, allows a different question: when you flatten the highlight reel, who's actually left standing?

Wit Studio Leads on a Three-Show Sample, and That Caveat Matters

Wit Studio posts the highest average in the catalogue: 8.17 across three ranked shows, spread ±0.58. The flagship is Vinland Saga (2019), which the Codex scores at 8.88 — MyAnimeList puts it at 8.78 — with character writing at 9.2 and story at 9.0 carrying the weight. Thorfinn's transition from revenge engine to farm-arc protagonist is the kind of long-form character arc the rubric is built to detect, and the production work on the prologue episodes under director Shuuhei Yabuta gave Wit a calling card that still defines the studio's identity.

Three shows is three shows. A studio that has only ranked three titles hasn't been tested the way Madhouse has been tested. The 8.17 average is real and the spread is reasonable, but it's the kind of number that can move half a point with a single weak entry. Take Wit at the top of the table; don't take Wit as a settled verdict over a studio with six times the sample size.

Tokyo Movie Shinsha and the Cultural-Impact Tax

Tokyo Movie Shinsha lands second at 8.09 across three shows, with a wider ±0.72 spread. The flagship case is Ashita no Joe (1980) — 47 episodes, shonen, Codex 9.02, MyAnimeList 8.83. The score is propped almost entirely by cultural impact at 9.8, themes at 9.4, and character at 9.5; the animation criterion comes in at 8.0, which is roughly what 1980 limited animation can support before the rubric stops grading on a curve. The Codex argument for watching it in 2024 leans on exactly that asymmetry: a show whose runtime is justified by every axis except the visual one.

TMS's average benefits from owning a piece of medium history. It also illustrates why averaging works as a test — a studio that produced one of the most culturally heavy shonen ever made and only two other ranked titles is going to look different on the leaderboard than a studio with eighteen entries and inevitable variance.

Madhouse Holds, and the Sample Size Is the Whole Argument

Madhouse averages 8.06 across eighteen ranked shows, spread ±0.71. That's the number that matters on this leaderboard. Wit's 8.17 is built on three shows; Madhouse's 8.06 is built on six times that catalogue and still lands within a tenth of a point. The flagship here is Monster (2004), 74 episodes of Urasawa adaptation, Codex 9.24 against a MyAnimeList 8.89, with story at 9.5 and character at 9.7 doing the heavy lifting. The Codex verdict on Monster treats it as the seinen benchmark for runtime justified by character work, and the rubric reads it the same way: a thriller that earns every one of its 74 episodes on Tenma, Johan, and Nina rather than on plot mechanics.

The Madhouse case isn't that any single show outscores Wit's flagship. It's that the studio sustained a top-tier average across eighteen entries spanning decades, directors, and demographics. That's the closest thing the catalogue offers to a structural argument about a studio rather than about a season.

Bones Ties Madhouse on Average, Loses on Spread

Bones matches Madhouse's 8.06 average — across six shows, with a wider ±0.98 spread. The flagship is Fullmetal Alchemist: Brotherhood (2009), 64 episodes, Codex 9.25, MyAnimeList 9.11, with story at 9.5, cultural impact at 9.5, and character at 9.4 all clearing the bar. It's the highest-scoring show on this leaderboard's flagship list, and the studio's strongest single argument that shonen can hit every criterion when the source material and the adaptation align.

The ±0.98 spread is the catch. Bones has a higher ceiling than its average suggests and a lower floor than the headline number implies; the Mob Psycho 100 season-two review sits near the top of that range. Six shows is enough to start trusting the average, not enough to make the variance look like a feature.

Production I.G, MAPPA, and the Middle of the Table

Production I.G averages 7.73 across seven shows with the tightest spread in the upper bracket at ±0.56 — including Usagi Drop (2011), Codex 8.32, MyAnimeList 8.31, an 11-episode josei whose character writing at 8.7 carries it. MAPPA averages 7.72 across five shows at ±0.80. The MAPPA case is volatile by construction: the catalogue includes work like Dorohedoro, whose 7.51 the Codex treats as a beautiful almost-masterpiece stranded mid-sentence. Artland (7.70) and Tezuka Productions (7.67) close out the ≥3-show top eight.

The Consistency Bracket

The most consistent studios with at least four ranked shows aren't on the leaderboard above. David Production posts a ±0.36 spread, Studio Comet ±0.37, and Gallop ±0.39. The averages are lower than the top bracket — none of these are flagship-tier studios in the Madhouse sense — but the variance is. The rubric reading is straightforward: these are studios you can predict. A new David Production show is going to land within a third of a point of the last one, which is a different kind of reputation than Wit's or Bones'.

The Counter-Argument: Averaging Punishes Ambition

The strongest objection is that a flat average rewards safe catalogues and punishes studios that take swings. A studio with one 9.3 and one 6.4 averages to the same number as a studio with two 7.85s, but only the first one made a great show. By that logic, Madhouse's average understates the studios with high-variance catalogues, and the leaderboard becomes a measure of curation rather than achievement.

The rubric reads it differently. The Codex score is itself a six-axis average — a great show is one that performs across all six criteria, not one that spikes on cultural impact and craters on world-building. A studio average is the same logic at the catalogue level. If swing-for-the-fences output produces one masterpiece and four mediocrities, the rubric is correctly reporting that the studio's typical output is mediocre. The flagship still gets its individual score. The studio doesn't get credit for it twice.

Wit leads the table on three shows; Madhouse owns it on eighteen. Bones matches the average but pays for it in spread, and the consistency bracket — David Production, Studio Comet, Gallop — is its own argument about which studios you can trust without checking the season. The leaderboard isn't the order the discourse remembers, which is the entire point of running the numbers.