Fun With Numbers: Short-Term Versus Long-Term MAL Popularity

Back in late April, I was taking a look at some available numbers which may indicated casual interest in a show, including myanimelist statistics, for the disk sales and print bumps of series aired in the latter half of 2013. In order to get the mal values, I pulled them from the site a couple of days before I posted the relevant articles. One of the comments on this article raised a very legitimate question; how did I know that these values were consistent with the ones a series had in midseason, around the time we would hope to use them to predict print boosts and disk sales before they happened. That was a key assumption – that series’ relative popularity and ratings would remain constant after the end of the season. This turns out to be a bit of an oversimplification, especially for older seasons of anime where things have had more time to change.

Using fantasy anime league data, I was able to go back and look at both the popularity and ratings a show had at the end of a season (i.e. prior to the beginning of the next season). Many seasons of data are available, but I have chosen to focus on 4: Fall 2013, Spring 2013, Fall 2012, and Spring 2012. Note that Summer/Winter seasons are not tracked by fal. This post focuses on my analysis of the evolution of the popularity numbers (collected here) as compared with values taken in late July, 2014. Namely, I want to know if series exhibit significant postseason changes relative to their peers and, if so, which series do so most prominently. I’m still playing with ratings data, and will address those in a separate post later.

Note that Kyoukai no Kanata and Kill La Kill were excluded from the Fall 2013 fal season for understandable competitive balance reasons. Their exclusion is nonetheless somewhat frustrating, as they would have been interesting to look at through this context.

For the original work that prompted this study, I used models which involved singling out the top 5, 10, and 15 shows in a given statistical category. A simple way to check whether the effects of time on popularity on those analyses are significant is to check and see whether any given show’s change relative to its peers was enough to have it switch from one of those groups to another (i.e. something rising into or falling out of the top 10).

In total, the 4 seasons measured show 15 total shows changing groups, affecting the season’s 5-10-15 groupings in the following ways:

Fall 2013:

-Outbreak Company (originally top 5) and Kuroko no Basket’s second season (originally top 10) switch groups.

Result: Top 5 analysis is affected.

Spring 2013:

-Suisei no Gargantia (originally top 5) and Date A Live (originally top 10) switch groups.

Result: Top 5 analysis is affected.

Fall 2012:

-K (originally top 5) and Sakurasou no Pet na Kanojo (originally top 10) switch groups.

-Code Breaker (originally top 10) and Shinsekai Yori (originally top 15) switch groups.

Result: Top 5, top 10 analyses affected.

Spring 2012

-Kuroko no Basket rises from top 15 to top 5.

-Tasogare Otome x Amnesia drops from top 5 to top 10.

-Medaka Box+Mysterious Girlfriend X drop from top 10 to top 15 slots.

-Acchi Kocchi rises from top 15 to top 10.

-Eureka Seven season 2 replaces Uchuu Kyoudai in the top 15.

Result: Top 5, top 10, and top 15 analyses all affected.

Though the effects of time are less severe for more recent seasons (possibly because series with particularly strong reputations continue to add fans rapidly while those with middling reputations do so at a much slower rate), they do still play a role at the top of the list. And in older seasons, they start really messing with the original end-of-season groups. It’s definitely not okay to assume that myanimelist rankings relative to other contemporary shows tend to remain constant over time.

All this doesn’t necessarily mean that the numbers themselves aren’t useful, though. In fact, the modern popularity numbers can be combined with the end-of-season numbers to yield an additional way of looking at the merits of a given series. There’s a school of thought that goes that a lot of anime become popular simply by virtue of being new, despite providing replacable or subpar content when compared to critically acclaimed “backlog” titles. If this were true, one of the ways it might manifest itself is in series getting relatively more attention after they spend significant amounts of time off the air.

Getting a percent increase in popularity was a fairly simple matter of dividing a series’ current totals by its end-of-season totals. That value isn’t a good way of comparing shows between seasons, though, as some shows have had more total time to gather fans than others. In order to get a value that could be compared between seasons, I computed a z-score based on how much a show’s rate of popularity increase outpaced others from its same season. If a show’s rate of adding fans was a standard deviation better than the season average, it had a z-score of +1. If a shoe’s rate was half a standard deviation *worse* than average, it had a z-score of -0.5. The series in the sample with exceptionally high or low z-score values are listed below (exact values are in the source doc linked to in the second paragraph, if you’re curious).

Series with a z-score of 1 or higher (most positive to least positive):

Kuroko no Basket
Shingeki no Kyojin
Date A Live
Sakurasou no Pet na Kanojo
To Love Ru (season 3)
Accel World
Magi (season 2)
Kuroko no Basket (season 2)
Log Horizon
Psycho Pass
Girls und Panzer
Hajime no Ippo (season 3)
Nagi no Asukara
Yowamushi Pedal
Toaru Kagaku no Railgun (season 2)
Hataraku Maou-sama
Jojo’s Bizarre Adventure

Series with a z-score of -1 or lower (most negative to least negative):

Lychee Light Club
Busou Shinki
Galilei Donna
Kuro Majo-san ga Tooru
Hidamari Sketch (season 4)
Sekai de Ichiban Tsuyoku Naritai
Samurai Flamenco
Furusato Saisei: Nihon Mukashi Banashi
Hayate no Gotoku (season 3)
Robotics Notes

There’s a couple different things these numbers seem to represent. They seem to be helped by a series’ continuation; OreGaIru and Hataraku Maou-sama got into the top group with 1-cour seasons and nothing since, but their disk sales peg them as likely candidates for continuation in the near future, and most others in high plus territory either had a second cour or a second season. Not all second cours are beneficial, though. Robotics Notes and Samurai Flamenco were two shows that ended up at the bottom of the pack despite theirs.*

Interestingly, the highest z-scores don’t just fall to series that were originally popular – Kuroko no Basket, the recipient of the biggest gains percentage-wise, was 13th in popularity at the end of its season, and would be 3rd among its peers if you compared them today. This may be partially due to the fact that it did have more time to distance itself from the pack, as discussed above, and partially due to the fact that it was, for about a year, the only series targeting female sports anime fans as effectively as it did, making it difficult to replace with seasonal fare than shows from other genres.

This stat is interesting to me, and I’ll definitely be comparing it with ratings as I look at the time evolution of that stat as well. Anything that results in additional insight is generally something worth putting time into.

*The specific contents of the second cour might not be the only reason why they are where they are, but I doubt they helped them get positive word of mouth.


One thought on “Fun With Numbers: Short-Term Versus Long-Term MAL Popularity

  1. Pingback: Fun With Numbers: Short-Term Versus Long-Term MAL Rankings | Animetics

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s