About the Math
The Core of the Charts is made up of data from three sites: Amazon, Barnes & Noble, and Borders.
Once a week, I visit each site to check their Manga categories, and I sort the search results by ‘bestselling’. [The links above will pull up exactly that.]
I then click through, page after page, and type the titles into a spreadsheet in the order that they are ranked on the sales site. [this is the hard part]
And once I have a full list, I assign points to the books depending on how highly they rank. Add up the points each title earns (and add on similar data from a half dozen second-tier sales sites) to get a composite score, and there’s your ranking.
In concept, it’s that simple.
##
not all sources are treated equally:
Class 5
These two sources have one (annoying) thing in common: their book listings have a Comics/Graphic Novel Category, but do not break the listings down any further than that. So, while a researcher knowlegeable in manga and related books can certainly go through dozens of pages to find manga mixed in here and there, it’s almost not worth the effort. [and yet, I make that effort…] The current iteration of buy.com has another limiter: their bestseller list for ‘graphic novels’ only goes out 10 pages (20 per page) — when I can’t find enough manga mixed into the GNs to fill out a top 100, I have to stop.
Given these limitations, I’ve discounted these two sources — such that a book that garnering #1 honours only merits 5 points, and I only look up a [maximum] 100 titles from each.
Distro:
[yeah, those poor little numbers seem all but lost in the bottom corner — that’s because all charts in this post were made to the same scale, so very soon you’ll see that yes, in fact the numbers here are kinda insignificant]
Class 4
Hastings.com, Powells.com, and DeepDiscount.com
These are still small fry, relatively speaking, and so I only consider and score a top 100 to match the Amazon spot checks (and the two Class 5 sites) — but they each have a usable manga category listing [*angelic chord, with sunbeam*] and so we’re able to stretch a search just a bit further into the backlist:
For each of these three sites I log and score 100 titles, plus 30 more. The extras at the end are only scored at one-tenth of a point each (minus another really miniscule fraction as we go down the list, to aid in breaking ties) — but, given the quality of the source data the #1 ranked title gets 10 points.
The flat line hugging the bottom of the graph at the right are the ‘extras’ — not quite zero, as each extra is scored at one-tenth point. —It’s not enough to move an individual title more than one spot up or down in the rankings (if that) but it does add up for the series ranking, and this +30% a great way delve deeper into the mid- and backlist.
Class 3
Not quite in a class by itself [I’m currently re-evaluating Chapters of Canada for inclusion in these charts] but still not-ready-for-prime-time and also just a regional bookstore (and marginal online) player — that said, I like the Books-a-Million site, it gives me results I can use, and also reflects a portion of the business that isn’t covered by the Amazon/B&N oligarchic hegemony.
I double up on BAM, looking into a top 200 (and out an additonal 33%, another 70 or so titles scored at just one-tenth of a point) and the scores are also doubled [#1=20 points] though there is a slightly different distro:
Class 2
Amazon doesn’t just get counted twice, I actually check in to Amazon five times a week. The first four are quick checks: The 100 Manga Bestsellers for a given [mostly random] hour on each of four consecutive days — typically Thursday-Sunday each week, though it can shift depending on my work schedule.
Theoretically: For each of Amazon’s Hourly lists, #1 should get 10 points, with a now familiar slide down from there.
Since Amazon doesn’t know any actual books from its asshole elbow and relies on user tags, half-assed keyword searches, and guesswork advanced hueristics to determine category placement and search results, often I have to discard 15-25% of Amazon’s posted list because it’s just not manga. Sometimes it’s not even comics, let alone manga related.
So a chart would look like this:
except the Hourly Bestsellers are a top 100 — no more, no less — so any skipped titles affect our imputs: the area shaded in green on the chart above represents the 25% [max] of each list that wasn’t manga and so can’t be scored. Actually, it’s 4 charts, so let’s multiply that out
And suddenly we’re looking at up to 4 points dropped. Big deal, right? 4 whole points?
Actually, that can be the difference between #10 and #11 — between making the chart everyone reads and just missing it, to end up as background data.
What’s a math geek to do?
Well, in this case, I overcome Amazon’s deficiencies by running the Hourly Bestsellers [as scored above] through their own mini-spreadsheet to get a transitional ranking, usually ending up with 150-190 titles (though I’ve made allowances for up to 250 unique listings using this method) and scored like this:
with a #1 equal to 30 points (there are 4 source charts rolled into this one) and an all-new, much improved points distribution that approximates what would have been the sum of four top 100 charts [though discounted at the top by 10 points, because Amazon pissed me off, making me do extra work]
Class 1
Amazon, Barnes & Noble, and Borders.
Again, these are the biggest sites, my best sources, and [well, Borders gets a pass…] the only online retailers of books doing Hundreds of Millions of dollars in biz. It’s the big time; this is main show.
Amazon was counted above (4 times!) with their hourly bestsellers, but I also load up their manga category listings once a week for a deeper look into how the wanna-be-monopolist is doing with manga sales. Along side, I load up B&N and Borders listings for the same books, and together, these 3 sources make up half the books (52% by number plus or minus 1%)(plus or minus because the total number of books can vary due to insufficient data from sources) but 80% of the total points (81.675% as weighted – for those math freaks who really must know).
I pull a top 300, each, from The Big Three — +33%, another 100 titles at one-tenth point each, to track the head of the long tail — and the results are also given greater weight, given the greater sales volume of these sites. #1 scores a full 60 points, & the following distribution looks something like this:
##
Assuming all sites were not only comparable, but interchangable — with the same titles in the same order on each ‘bestseller’ list — then given the points & weighting as assigned above, the chart would look something like this each week:
[itsonlyamodel]
[/itsonlyamodel]
[please note the extra zero added onto the y-axis; now we’re talking hundreds of points, not dozens.]
Of course, the differences are the point of the exercise.
instead of the same 400 titles every week, the spreadsheet tracks many, many more — and the actual distribution looks like this
[from the charts as posted, week ending 11 July 2010]
you might have to scroll right for a bit to catch all that.
So, even with very crude constructs to shape it, once the real world data is shoe-horned into my spreadsheet we get results that seem to match much more complicated models of the retail market. If anything, I assign too much weight to the midlist titles (say, from #30 on down to #500 or so) but this ‘error’ or distortion is fine by me, as it helps differentiate and sort a lot of same-same-y seeming titles past Naruto, Bleach, and Maximum Ride.
##
For me [at the moment] this is just a fun hobby. (Yes, math is fun.) (…back off)
So if you’re about to send me an email objecting to this point or that about my method and why I’m wrong, wrong, wrong please consider:
- I do this part time.
- I do it for free.
- I’m at the mercy of my sources. Don’t complain to me: take it to Amazon.
- My charts are not meant to be authoritative or even correct. I compile comparative rankings of manga titles and series based on online sources (as discovered, and publicly available) and while I hope to approximate actual sales numbers, I don’t have access to that data and never will. These aren’t even estimates, as I only know that “title A beat title B” for a given week, and only if the sales sites I reference are being honest with their reporting. I’m doing the best with what I have, folks.
- That said, this twice-removed approach brings up some really interesting starting points for discussion, analysis, and debate.
##
Additional inquires about the charts can be emailed to me: the contact info is lying around somewhere and genuine questions will be answered with genuine responses.
This is the latest iteration of a long experimental process; additional info on methods and procedures can be found here[v2] and here[v1]
