How to Read a MyAnimeList Score Before You Trust It

Attack on Titan’s first season scores 8.58 on MyAnimeList. Season 3 Part 2, ten episodes from much further into the same story, scores 9.05, and about 1.3 million fewer people voted on it. The audience may have changed, and so may its reaction to the story. These totals alone cannot tell us how much either contributed to the score gap. A MyAnimeList score is a lightly weighted average of 1-to-10 ratings from the members who chose to watch and rate that one entry. Before you let it decide anything, check how many people scored it, which season it belongs to, and whether it has finished airing.

Seven scores side by side

These figures were read from each entry’s MyAnimeList page on 14 September 2026. They will have moved by the time you read this, but the relationships between them change slowly.

“Members” counts everyone who added the entry to a list, including people who dropped it; the last column shows what share of them left a rating MyAnimeList accepted.

What the number is

MyAnimeList publishes its formula: Weighted Score = (v / (v + m)) × S + (m / (v + m)) × C. S is the entry’s plain average, v is how many users scored it, m is the minimum number of scorers an entry needs before it shows a score at all, and C is the mean score across the whole database. Put simply, when few people have voted, the result is pulled towards the site-wide average, and when hundreds of thousands have voted, it is essentially the plain average. Entries below the minimum show no score.

Two filters sit in front of that. MyAnimeList removes scores from accounts it judges illegitimate, and once a series completes, it discards scores from users who watched less than a fifth of it. So the number you see is a smoothed average of people who opted in and got a fifth of the way through, and the opting in is where most of the distortion comes from.

Sequels are rated by the people who stayed

Attack on Titan Season 3 Part 2 follows 49 earlier episodes. Its viewers are likely to include many people who enjoyed the earlier seasons enough to continue, while people who dropped the series may never rate the later entry. That selection can contribute to a higher sequel score. But the similar shares of members who rated each entry, 70 and 69 percent, do not establish what caused the score gap. Viewers may also prefer the later season itself; aggregate counts cannot separate those explanations. Seasons, cours and episode counts explains why one story ends up as several entries in the first place.

The same climb shows up in the AniList ratings Fantera displays. HAIKYU!! rates 8.4, its second season 8.6 and its third 8.7. Mob Psycho 100 goes from 8.4 to 8.7 across its three seasons of 12, 13 and 12 episodes. Bleach, 366 episodes from 2004, sits at 7.9, while its 13-episode continuation BLEACH: Thousand-Year Blood War from 2022 sits at 8.8.

It is a tendency, and Frieren breaks it. The 10-episode second season scores 8.84 against the first season’s 9.25, from under a third as many voters. A later season can receive a lower score, but these totals do not tell us whether the same viewers rated both entries. Read their reasons before interpreting the change. For deciding whether to start a franchise, though, the first entry’s score is the one that describes an audience like you.

Airing shows and long runners

BLEACH: Thousand-Year Blood War - The Calamity is still airing as this is written, with ten episodes scheduled from 25 July to 27 October 2026, and it already scores 9.08. Only 19 percent of its members have rated it. The people scoring the fourth part of a revival week by week are the franchise’s keenest viewers, and the fifth-of-the-series rule only applies once a show completes. Treat any score on an airing show as provisional, and expect it to move once the season ends and slower viewers catch up.

Long runners filter in the other direction. Naruto: Shippuden runs 500 episodes, so a fifth of it is 100, and under MyAnimeList’s rule anyone who quit before episode 100 is left out of the score. Gintama’s first series, at 201 episodes, works the same way. What remains is the verdict of people who committed to a very long show, which tells you little about whether the first 20 episodes will hook you. Look for opinions about the opening arc instead.

When few people voted

Gintama the Movie 2026: Yoshiwara in Flames scores 8.81 from 3,661 voters out of 15,990 members. The formula nudges a small sample towards the site average, but it cannot change who is in the sample, and a new Gintama film is watched almost entirely by people who already love Gintama. The same holds on a larger scale for Gintama: THE VERY FINAL, the 2021 film that closes the series, which scores 9.05 from 89,008 voters. Both numbers are honest fan surveys. Neither tells a newcomer much about whether to start.

As a rough rule, a few thousand voters describe a community and hundreds of thousands describe a broad audience. In between, read one positive and one critical review that explain their reasons.

How this differs from Fantera's numbers

The star rating on Fantera comes from AniList, a different community with its own habits. Frieren shows 9.1 on its Fantera page against 9.25 on MyAnimeList, and neither figure is wrong: different people, different sites. AniList reports its score out of 100, and Fantera divides it by 10 for display.

The match percentage on a Finder card is another kind of number entirely. It measures how closely a pick’s tags resemble your anchor’s, blended with how strongly AniList users recommended it from that anchor, and it is rescaled and clamped so that no result shows below 40 percent or above 99. An 82 percent match means “made of similar things”, while a 9.05 means “the people who rated it liked it”. A show can sit high on one and low on the other: Wandering Witch: The Journey of Elaina is Frieren’s second-closest match at 81 percent and carries an AniList rating of 7.4. How Fantera recommends anime walks through that calculation.

So when you are choosing tonight, compare a first season with other first seasons, ignore gaps under 0.2 between shows with hundreds of thousands of voters, let airing shows finish before trusting them, and read reviews whenever the voter count is small. Still picking a first show? Where to start with anime beats any ranking.