Continuing on from my previous post on the notable changes of series popularity over time, I did some similar work on rankings. The shows analyzed are the same (all those tracked in fantasy anime league from Spring and Fall of 2012 and 2013), only this time I’m looking at the evolution of rankings; how series swap places in the rankings over time and which ones rise and fall most significantly. The data used this time around can be found here.
The first question here is how much the changes in rankings from end-of-season to the present time has affected the order of the shows (this has the potential to affect my previous analysis of them as a predictive tool). Unlike popularity, where the big changes in order, while present, seemed to take time to really shuffle the ‘original’ end-of-season order, changes in ratings get fairly significant fairly quickly. The most recent season, Fall 2013, contains 7 shows that switched groups between the top 5, 10, and 15 shows, and only Spring 2013 contained less than 5 swaps.* Essentially, the part of the analysis that used the modern numbers doesn’t reflect what things were like at the time.
Why do rankings change so much faster than popularity totals? One possible explanation is that it’s fairly easy for small numbers of postseason viewers (~1000 users or so) to cause the numbers to fluctuate for less popular series. Another stupidly obvious explanation pops out when one looks at the series showing exceptional degrees of change against their peers (Z-scores again):
Series with a z-score of 1 or higher (most positive to least positive):
Nagi no Asukara
Gundam Build Fighters
Rock Lee no Seishun Full-Power Ninden
Girls und Panzer
Jojo’s Bizarre Adventure
Series with a z-score of -1 or lower (most negative to least negative):
Eureka Seven (season 2)
Tonari no Kaibutsu-kun
Chuunibyou Demo Koi ga Shitai
There are too few series with exceptionally negative Z-scores to say too much about what they mean, but every single series with a positive score has one thing in common; they weren’t finished at the time. Some just had a few episodes left to air (Girls und Panzer, AKB0048) while others would run for 3 or more additional cours (Uchuu Kyoudai, Shirokuma Cafe, Rock Lee), but all were missing what’s presumably a rather significant bump in rankings from the ratings of people who wait until the series is finished to give scores. And y’know, sometimes the second half of a series actually is the better half, though I always take those sort of claims with a grain of salt.** Given that, one might hope it would be possible to maybe salvage something applicable to print bumps using modern data for 1-cour series only, but that fact that chuu2koi and Tonari no Kaibutsu-kun both made it to the negatives means there’ll be issues even then.
Even more so than popularity numbers, ratings seem to be tricky to unpack. I could make an argument for modern ratings actually being better as an indicator of contemporary interest than contemporary ones; with the largest positive changes almost exclusively relegated to 2-cour or longer series, it’s almost certain that contemporary ratings leave out a very important demographic. Both sets of numbers have some obvious limitations, which is the most important thing to keep in mind.
At the very least, the end-of-show bump seems to be a potential future topic for study, one which I might look at if I get back to this data at some point.***
*That season was almost comically static with time. The top 6 stayed precisely pat and the only swap in the top 9 was Date A Live (originally #8) passing up Henneko to move up to #7.
**For all the usual reasons pertaining to subjective qualities, and more besides.
***Right now, I’ve got other, funner stuff that’s on my plate; revisiting License Effects, the Sequel Equation in general, late-90s timeslot exploration, analysis of a treasure trove of manga distributor print claims via mangahelpers, and how the US manga print market went from bottom-heavy to top-heavy. That’s pretty much a full-course meal.