How popularity is calculated

# How popularity is calculated Every series has a popularity rank, shown as something like _5th most popular overall_ and _the most popular manhwa_. Here's how we work that out. ## TL;DR We look at how widely a series is followed across the big manga databases (**AniList**, **MyAnimeList**, **Kitsu** and **MangaUpdates**), plus how many **MangaBaka readers** keep it in their own library. The tricky part is that every one of those sources counts popularity _differently_, and on _wildly_ different scales, so you can't just add the numbers up. Instead we rank each series _within_ each source, then fuse those ranks into a single score. That gives one clean list where every series has its own spot, recomputed daily, with history kept so you can see what's trending. ## Why we can't just average the numbers Each source has its own popularity signal and its own pool of titles. One might rank a few thousand series, another tens of thousands. Being near the bottom of the small list is _nothing like_ being near the bottom of the big one, but the raw numbers don't know that. So if you just average those raw positions, the scales fight each other: huge clumps of titles end up tied at the same score, and series that happen to sit on a smaller source get shoved way up the list for no good reason. You get thousands of ties and an order that doesn't really mean much. ## How we actually do it We use a well-established method called [Reciprocal Rank Fusion](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/reciprocal-rank-fusion) (RRF). Three steps: 1. **Rank inside each source.** For every source, sort its series by that source's own popularity signal. #1 is the most popular there, and on down the list. 2. **Turn each rank into a score.** A high rank is worth a lot, and it tapers off smoothly as you go down. 3. **Add them up.** A series' score is the sum of what it earned from every source it shows up on. Higher total = more popular. The formula for one series is just: ``` score = sum, over every source the series appears on, of: 1 ───────── 60 + rank (rank = the series' position on that source, 1 = most popular) ``` The `60` is a standard smoothing constant (the conventional RRF default) so the very top isn't _absurdly_ spiky. Because step 1 turns everything into ranks first, the mismatched scales stop mattering: being #100 on a small source counts the same as #100 on a big one, and no single source can run away with it. To get the number you see, we sort every series by that score. The position in that sorted list is the **rank**. We do it twice: once across the whole catalog (the _overall_ rank) and once within each format (the _type_ rank). ## Why coverage matters A series only earns points from sources it _actually_ appears on, so the more sources track it, the more scores stack up. That's on purpose: a title that's well known _across_ the manga world should beat one that's only big in a single place. Quick example: being #1 on _one_ source is worth about `1 / 61 ≈ 0.016`. Being #100 on _four_ sources is worth `4 × (1 / 160) ≈ 0.025`, which is higher. Broad presence beats a single strong showing. The MangaBaka library counts as one of those sources, it's just our own: how many readers actively keep a series in their library (we leave out the ones marked _dropped_). It's small today but grows with the community. ## One ranking, no ties Sorting by score gives every series a distinct position. When two series land on the same score we break the tie consistently, so the final ranking has no shared spots or gaps. That's why a badge like _5th most popular_ can link you straight to where the series sits in search, instead of a vague ballpark. You'll see two numbers on a series: - a **global** rank across the whole catalog, and - a **type** rank within its format, so manga is ranked against manga, manhwa against manhwa, and so on. That's the _"5th most popular overall, most popular manhwa"_ kind of split. By default the badge shows up when a series is in roughly the **top 1,200** of either its global or its type rank. ## Trends over time We take a snapshot of popularity every day and keep the history, so on top of "where does it rank right now" we can show how it's _moving_: climbing, sliding, how fast, and the change over the last day, week, month, all the way out to a year. It's computed daily for now. Depending on how the data behaves we might switch to weekly later, but we'll see. ## Caveats & things to keep in mind > [!IMPORTANT] > Popularity is _not_ a quality score. A high rank just means a lot of people are reading and tracking a series, not that it's any good. Ratings are a separate thing entirely. A few more things, so we're upfront about what this does and doesn't do: - **It's coverage, not merit.** A niche or brand-new series can rank low simply because there isn't much data on it yet, not because it's bad. - **We're only as good as our sources.** The rank is built from each source's own numbers. If a source is stale or off, that flows through. We refresh regularly, but we don't second-guess what the sources report. - **Every source is weighted equally.** We don't currently treat one database as more authoritative than another, each contributes on equal footing. - **Our own reader signal is still young.** MangaBaka library counts are small next to the established databases, so for most series they nudge the result rather than drive it. That'll change as the community grows. - **Older trend history is reconstructed.** We only started taking daily snapshots recently. History further back was rebuilt from each source's change records, so older points on a trend are approximate, best-effort rather than an exact daily reading. Everything from when we started capturing forward _is_ an exact daily snapshot, and the long windows (like the 1-year change) only show up once a series has that much history behind it. - **It's relative and it moves.** A series is reranked every day, and its number shifts as the whole catalog and reading habits change, even on days nothing about the series itself changed. ## FAQ **Why isn't my favorite series ranked higher?** The rank rewards being known _broadly_. A series that's huge on one platform but barely present elsewhere will sit below one that's solid across several. And remember it's popularity, not quality, so a great but lesser-known series can absolutely rank low. **Why did a series' rank change when nothing about it changed?** Because it's relative. We re-rank everything daily, so when other series rise or fall, positions shift around them, even for a series whose own numbers held steady. **Is this the same as the score / rating?** No. Popularity is _how many_ people follow a series. The rating is _how highly_ they rate it. Two different questions, two different numbers. **Does my library affect it?** Yes, a bit. Reader libraries are one of the signals (dropped series don't count). It's a small part of the mix today and grows as more people use MangaBaka. **How often does it update?** Daily, for now.