If you’re not already aware, Seasonal Prattle messes around with anime statistics quite a bit. We love seeing how airing titles are being scored across various platforms, their changing drop rates and holds as a season unfolds. We even had at one point a segment attached to our typical entries dedicated wholly to quick stats.

That is, until we realized how incomplete they are.

Some swear by stats. They look at online scores and evaluate a series’ quality primarily, or worse, purely off that value – an approach that’s shaky for various reasons. The most prominent among them being simply how arbitrary scores can be in the community.

What the value assigned to a series actually means will vary from person to person. For example, It’s common for a seven on a frequently used site like MyAnimeList to actually feel like an average rating given its high usage – despite that seven meaning “good” as the site’s scale intends. On the same hand there, series scored with a five or a six generally are disregarded and referred to as “bad”, although their ratings are suppose to translate to average and above average.

This alone can complicate a given work’s collective reception since not everyone perceives and assigns scores the same. Sometimes a five really does correlate to its intended value of average. Sometimes a ten really is a work that someone felt was a masterpiece to them and not just one they simply liked more than usual. But sometimes a show that someone felt was “okay” gets thrown a seven. Sometimes a show someone didn’t like a little gets thrown a one. The problem is it’s virtually impossible to know every time, and thus, it can be a bit muddy getting a good community pulse on a show.

With that said, for added accuracy and comprehension we’ve turned towards natural language processing software in recent weeks in addition to the typical statistics. For those who aren’t familiar with it, it’s also called text analytics or computational linguistics (this kind of software has both open free versions so you can google into it and give it a shot if you were curious).

It’s basically just social listening software with more direct reach beyond the social platforms to identify and analyze opinions in text on the internet (or subset of the internet like a particular forum) with a sentiment output attached (generally good, neutral, positive on a sliding scale).

That output better aids us in the direction of a show’s conversation and allows potential reinforcement to what typical stats might suggest (or not).

So instead of looking at let’s say Planet With from this season, seeing its score on MAL (at the time of writing it’s a 6.76), cross-checking its statistical history with similar platforms and their corresponding reviews, touching base with key critics and come to a conclusion – we can also now systematically evaluate the greater conversation’s mood in real time to get a better idea of where the consensus is as well.

This approach of course isn’t perfect, but the accuracy has been pleasing. We feel it’s a valuable step forward in determining where the community truly stands versus what’s just noise and false signals.