Six months ago, I took my amazon data and banged out rough sales projections for the first half of 2016. Just recently I found the time to do the same for when the region in question made an unusually high number of terrible decisions motivated by the hate that apparently fuels a depressing number of people.

Disclaimer: In this post I’m going to do some funky stuff with data that research has led me to believe I can reasonably do, plus one very uncertain gap-filling reach. BUT I don’t have any inside information and I could make mistakes here in a number of ways.

Note: ‘release’ can mean multiple versions of the same anime content, i.e. BD and DVD versions, as well as regular/limited/collector’s editions.

Formula Recap:

-Based on meticulous observation, a series with an amazon rank of R on a given day will sell S copies, as follows:

#10k-#40k: S=10-(R-10000)*(9/30000)
#2k-#10k: S=30-(R-2000)*(20/8000)
#1k-#2k: S=70-(R-1000)*(40/1000)
#500-#1k: S=220-(R-500)*(150/500)
#200-#500: S=250-(R-200)*(30/300)
#150-#200: S=800-(R-150)*(550/50)
#100-#150: S=1000-(R-100)*(200/50)
#50-#100: S=1300-(R-50)*(300/50)
#40-#50: S=1900-(R-40)*(600/10)

Note: Direct predictions for Only Yesterday on BD are difficult because of its extremely high ranks, but the best guess (setting all >#40 days to 1900 sold) is only off by about 2000 from the actual known week-1 total, so there’s that.

THE BIG FUDGE that’s going to come in in parsing this data is that I’m missing part of one week in September/October affecting 32 releases, which are missing either 7 or 4 days in the middle of the period. This somewhat affects their figures, but since the key factors influencing the calculation of the eventual total are the SECOND week figures, and those are more stable from day-to-day, we could potentially skate past the mishap without major issue. HOWEVER, those 32 releases account for about 160k copies projected over the year. If that’s increased by double as a result of the missing days, it adds about 5% to the 2.1 million copies projected to be sold. Not enough to completely remake the sample, but enough to shift narrow conclusions. Ultimately, given the factor-of-2 accuracy of the core algorithm, I don’t expect it to be a major factor. If that does happen to be the case, it sucks, I can’t do anything about it, and we’ll never know that it was. It wouldn’t be the first time this year my source data has been an imperfect representation of reality. Ultimately, when you’re analyzing frustratingly opaque markets, you work with what you have.

This process gave me projected 1-year totals for each release, assuming BD sales trends. Where applicable, I multiplied these BD figures by a factor of 3 for DVD releases. In case someone new is reading this, let me just say that this model of DVD release behavior, it is not an assumption. It’s a projection of characteristics of the market I have observed and have concrete evidence for, but extrapolated into murkier territory. Thus, it’s a statement that is neither flawless nor groundless to make. The model itself is only good to a factor of 2 after DVD corrections are factored in, so there’s a lot of range that should come in. On release didn’t definitely do better than another because it has a slightly higher projection, but something projected at 30k is beating something projected at 2, for damn sure.

—

Here are some points that arise from the dataset (ASSUMING IT’S SOMEWHAT VALID AND THERE IS ROOM FOR DOUBT ON THAT FRONT):

-Overall less money ($57 million vs $69 million) will be grossed on more releases (257 vs 217) compared to the first half. Almost all of this gap comes from within the top 50 [which contains 7 of the contaminated/gap-filled releases] ($39 million vs $49 million). The top 10 figure appears to have held steady, but that’s a somewhat apples-to-oranges comparison – the presence of the $140 Code Geass megabundle really obscures a lack of other peak performers. Also, 101 releases are projected to sell under 2000 copies, versus 56 in the first half. Essentially, the midsection held sorta ok but the lower and upper tiers were a bit unsteady.

-Within my sample, though, Only Yesterday slayed and is projected to be the best individual release of 2016 by a wide margin. It’s projected at 355k.

-The top of the sample is overall less interesting – a lot of re-releases or part X of the continuing battle/fanservice franchises (which seem to be on a roughly equivalent sales tier). Trinity Seven stands out as the relatively new TV series selling quite a bit. With BD and DVD combined, it projects to sell roughly 50,000 copies in its first year.

-I left out Final Fantasy Kingsglaive because it’s not tagged as anime on amazon so it didn’t come up on my initial pull, but the video game movie will almost certainly beat everything from both this and the first half. With 60k in the first week and 10k in the second, it would project at close to 600k total sales in its first year, for about 9 million dollars grossed. Count that and the two halves of the year come out almost even.

-Crunchyroll has over 1 million subscribers now, and if each of them chipped in $60 they’d have an annual war chest of 60 million. That’s a little less than half of what I predict all of 2016’s new disk releases are worth within the US. Not an apples-to-apples comparison because other CR markets worldwide and other countries have disk markets I know little-to-nothing about. Personally, I’m grateful that CR releases these numbers, even though it opens them up to some pretty wide-scale speculative data-mining which I would definitely do if I weren’t living in the land where the only thing premium sub was good for was Investor Z chapters (still not a terrible deal, truth be told). There’s at least two very clear roadmap to estimating the costs of their streaming licenses with publicly available data, and I lament the fact that neither is available to me personally at present. If anybody lives in the US and wants to be a gremlin for that, hit me up.