on the quest for some motivation

So the flavour of the day is Bad Fandom Stats, and it turns out that not only is centreoftheselights using a wildly inaccurate method of counting ship popularity during a given year for the Year In Review fandom stats, but also that nearly every column on the chart is a lie. @5ummit exploded the whole methodology issue last year, so I'll just link their post and dive into other stuff.

One would think, looking at a list of popular ships with their fandoms listed next to them, that the named fandom is simply the one within which the ship exists.

The ship at the top of that list is "Evan "Buck" Buckley/Eddie Diaz (9-1-1 TV)" and the fandom is listed as "9-1-1 (TV)"; that's pretty straightforward, there's only the one TV show and those characters are from that show.

But then we get down to ships with characters that exist in multiple versions or subsections of a canon which have their own fandom tags on AO3, and things start getting janky

One of the first things I noticed was weird about this year's chart (aside from the numbers themselves being just straight up wrong) was that the fandom for Kirk/Spock was listed as "Star Trek: Alternate Original Series (Movies)", the AO3 tag for the reboot movies. That felt wrong, because while I know the reboot movies are big on AO3, most of the K/S stuff I've seen recently has either been expressly original cast or not specific to any one cast or iteration of canon.

I thought that the list might have been saying that Reboot Kirk/Spock alone was big enough to make the list while Generic and Original Kirk/Spock were separate fandoms that hadn't gotten onto the list. That would be absurd for new fic count, but there are stranger things on this list and the methodology favours newer ships, so I went digging.

A search for Kirk/Spock fic posted in 2024 and a glance at the sidebar gives us this, out of 2,255 total fics

Searching for Kirk/Spock fics updated, rather than newly posted, in 2024 also puts the All Media Types and Original Series tags above the Reboot Movies.

But searching the entire unfiltered ship tag gets us this, out of 22,426 fics

This means two things.

1.) Despite the listed fandom on the year end chart, Kirk/Spock (and by extension, all the other ships from fandoms with more than one iteration of canon [and some Batman ships are on there, so we've got far more complex things than simply reboot movies and TV/anime adaptions of novels/manga]) is a generic ship tag.

2.) the fandom listed was taken from the unfiltered tag, not from this year's data.

And centreoftheselights confirms that this is indeed her methodology.

The fandoms are, quite simply, wrong.

--

Since we do not have access to her data and her methods are not replicable, we can't check how many of the ships might have been struck with a mismatch between which version of their canon is most popular overall vs this year.

I can't even be certain that there is a mismatch for Kirk/Spock. It's possible that the fics that actually make up her numbers have more reboot movie fic than otherwise. No one will ever know.

--

It is interesting to dig into the differences. I find it utterly fascinating that Kirk/Spock had a period that pushed the reboot movies to the top of the list overall and has since settled into Original Cast being more popular, all while the ships remained consistently popular enough to regularily end up on Top Ship Lists.

Centreoftheselights' data does not allow us to dig into those differences, or even to meaningfully speculate about them.

--

And yet more! As I explained in this reblog, the "type" column is not harvested from the data itself, but is a subjective interpretation of what centreoftheselights believes the characters' genders is or could or might be. So not only are all the columns wrong, they aren't even wrong in the same ways.

Over the whole sheet, the columns are: ⚠️Rank: Well, it accurately lists the order of the inaccurate counts, so I guess the column technically isn't inaccurate ⚠️Change: This is indeed the change in rank from last year's chart. It accurately lists the difference between two different inaccurate ranks ✔️ Relationship: accurate! This is the ship tag on AO3 ❌ Fandom: Inaccurate. Actually a top tag within the ship tag, not the fandom the ship is from ❌Works Gained: Inaccurate. see @5ummit's debunk ✔️Total Works: accurate! this is indeed the size of the tag on AO3 ❌Type: Inaccurate. OP's best guess, subjective interpretation, or headcanon ❌Race: Inaccurate. Same as above.

--

Tagging @olderthannetfic and @5ummit since you two have kind of been doing the heavy lifting on this one :)

Also plugging the more accurate chart by Randomist1031, which, in addition to having accurate fic counts, also lists fandoms by generic names rather than top tag, which increaes both accuracy and readability. https://archiveofourown.org/chapters/158271001

What do you think about this opinion from centreoftheselight?

https://imgur.com/a/0Wguz9t

I agree that there will always be inconsistencies and issues with any data collection method and that no method is perfect. This does not change the fact that the method they chose to use does not measure what they claim it measures, which is a pretty big issue.

They also say that every method will produce "the occasional strange data artifact" but completely omitting a dozen of the most popular ships (by any other metric) from your list of most popular ships is not just an "artifact." If a ship that gained 6000 works is not included at all and a ship that gained 2600 works ends up with a negative count, you should know right away that something is VERY wrong, all of your data is suspect, and that data should NOT be published, much less without serious warnings in big bold letters.

The other big issue with their methodology is that it's impossible to replicate because it relies on data only they have access to, which oddly they refuse to make public.

They're clearly aware that there are more accurate methods to determine popularity but they continue to use their current one because it's easy. I say if you can't commit to doing something right because doing it right would take more time and effort, don't do it at all.

I don't know if you're aware of this but the end of year AO3 stats posts are going around and one of the largest and longest running studies is extremely flawed (for example, reporting Dean/Cas had 2348 new works when they actually had 6456). Here's a post about it: https://www.tumblr.com/5ummit/738707388904898560/

--

Huh! I thought she was better at stats than this. We were on a panel together at one point, and she seems cool, but unfortunately, fandom stats are often not done all that well.

I know we all love graphs, but I want everyone to be critical of this stuff, including mine. I love Toast's work too, in addition to centreoftheselights, but I remember some old Wattpad stats where the methodology was to mark any work too incoherent to classify as gen... (LOL. Dude, have you seen Wattpad? Default gen, my ass!) There are tons of things wrong with every major fandom stats project or regular poster's work.

I haven't gotten on this soapbox lately, mostly because I've been too busy to run any of my own stats stuff heavily this year, but yeah... "Fandom stats" tend to consist of the same like... maaaaaaybe 4-5 fans, most of whom aren't statisticians.

I got invited to a con panel a few times based on... like... existing in public. And don't get me wrong: I do try to make sure shit I post is accurate and labeled as what it actually is, but I know I'm nothing hot when it comes to stats. I think I ask slightly smarter questions than usual and am willing to hand-count more things, but my actual "stats" are just "Here's the % of X. Here's the % of Y." and not a higher level analysis.

It's simply that the field is wide open with no competition. Aside from a tiny handful of repeat posters, it's just millions of randos grabbing the same few numbers from AO3 works search or filters and going "Gasp! Fandom has X% m/m!" (Ignoring that it's an AO3-only % and that everyone has access to this number and that it's a boring-ass thing to repost for the thousandth time.)

A new wrinkle is, of course, that if one uses one's personal account, one may have people muted. I suppose I'll have to get a new account if I want to be really accurate about stuff, though I think I currently only have 2 people muted and they're not prolific.

--

I don't want to put my own work down or that of the other people who post stats, but as audience members, we have got to get in the habit of reading the methodology section more carefully.

#if you're going to quote stats#check your sources#I'm a social scientist (qual) and I know not to do stats because I'm not trained in them#(I passed a quant methods course once but you can do that with zero stats knowledge. trust me)#with great power etc#fandom stats#fanfic and fandom

Yuuup.

I'm not going to stop posting little amateur charts of fandom things, but I want us to be very clear: I have looked into many of the people who routinely post fandom "stats", and basically none of us have any real qualifications, including me.

I don't think a degree is everything, but... Expertise is real. Studying how to study is real. Methodology is something professionals discuss a lot.

(Looking at you, horribly-designed fandom demographics surveys going around this month and every month!)

I make little charts of things because it's fun just like writing fic is fun, but let's not lose sight of the fact that it's easy to produce the equivalent of baby's first EFL songfic that hasn't even had spellcheck run on it.

I like to summarize the problem as:

"There's a pie chart so it must be true!"

#statistics #ao3 #fandom

Not really.

Part of being ethical in presenting data is helping your audience understand what the data actually say. It is blindingly obvious that the audience of the annual ship stats thinks they represent what fics got written that year. Since that's not the data being pursued, it ought to be in bold text at the top of every chart, not in an FAQ.

If the point is that net gain/loss is interesting, that should be telegraphed much more clearly.

Doing otherwise is tantamount to intentionally misleading the audience.

It is not mean to say so.

Thank you.

The data collection method is not simply a minor difference of opinion. It's not just that I personally would have done it a little differently. Their data straight up does NOT represent what it claims to represent (in their own words: "the most popular relationship tags on AO3 in the period Jan-Dec 2024"). More importantly and why I've been so "mean" about it, centreoftheselights knows this and continually chooses to obscure that fact because they know that what it does represent is (1) much harder to explain, (2) of limited interest and use, and (3) wouldn't get them as much attention and praise.

When your data is being seen and used by millions of people to do analysis and draw conclusions about the current state of fandom (there are people loudly and publicly celebrating "the death of destiel" because it's not on The List despite it still being the 11th most written ship in 2024), it's no longer just a fun personal project where you get a pass to do whatever you want, misinformation and lies be damned. As OTNF said, someone in that position has an ethical duty to present their data as clearly as possible and help their audience understand what it actually says, not what they want it to say.

If I've got to be a little over-the-top and "mean" to get people to pay attention and prevent the further spread of misinformation, so be it.

And while we're responding to accusations of unnecessary meanness in the tags, I want to address this one too since I've seen multiple people, including centreoftheselights, argue these points and it drives me insane:

Sticking to the same erroneous and misleading methodology year after year for the sake of consistency is not a virtue. Will people always find a way to misuse stats? Of course. That's why it's the duty of the person presenting the data to do everything in their power to prevent that misuse and misinterpretation upfront and centreoftheselights has failed in that duty miserably. It's abundantly clear that 99% of people who see these stats do not understand what they actually mean (and why would they when the author is obscuring the truth at best and lying at worst). Any sane and decent person who saw this much confusion and misinterpretation would be apologetic and change their behavior. Instead, they brush off people's valid concerns and double down.

I absolutely do not condone harassment. Nobody should verbally assault, or god forbid, dox anyone over fucking fandom statistics. But questioning an extremely influential "statistician" and holding them publicly accountable for spreading misinformation, especially when they continually refuse to take accountability and show zero remorse for their actions, is not harassment.

One of the largest problems with Centreoftheselights' ship stats is that the numbers are not a count of actual fics. The figures are better understood as a point based scoring system, and the ranking is a list of tags with highest scores.

The points are assigned thus: A new unlocked fic is created: +1 The tag is added to an old fic: +1 An old fic is unlocked: +1 A tag wrangling descision syns an old fic into the tag: +1 A new chapter is added to an existing fic: +0 A new locked fic is created: +0 A new unlocked fic is created and then locked: +0 A new fic is created and then deleted: +0 An old fic is deleted: -1 An old fic is locked: -1 The tag is removed from an old fic: -1 A tag wrangling descision de-syns an old fic out of the tag: -1

We can have conversations about which fics should be counted when we're talking about which ships are most popular within a timeframe. I like to use "created_by" searches in my posts about fic stats because those links are more futureproof, but an "updated_by" search, if you catch it close enough to the time, more accurately reflects fic-writing activity and quantity of new reading material. If we're doing a deep dive, we need to talk about which searches we're using and why.

We can also talk about if we should include archive locked works or not. I haven't heard any convincing arguments for excluding locked works from year-end stats presented to regular users with accounts, but there is absolutely value in doing both counts and comparing them.

But all of the above assumes that the number we end up with is a tally of all the fics on the archive that meet our inclusion criteria, and I don't think anyone is going to argue that a tag wrangling descision that re-sorts a ten year old fic reflects a ship being more or less popular this year.

--

The bad fic stats chart tells us that Jayce/Viktor "gained" 13261 works between August 2 2024 - July 29 2025.

But that "works gained" column is mislabeled, because it's actually that ship's score. The thirteen thousand is not a count of individual fics, and so we cannot go looking at those fics to determine anything about them.

When I tell you that there are 14,621 works in the Jayce/Victor tag that were created August 2 2024 - July 29 2025 and give you this link as my source, you can click on it and go read the fics. You can check the same link both logged in and logged out to see how many are archive locked. You can let me know if I'm being an idiot and searched the wrong time frame by accident. You can make observations about Jayce/Victor fic posted during that timeframe -- you can note, for example, that there are very few crossovers and that Trans!Victor is the top freeform tag. Maybe you're interested in how the percentage of Trans!Victor fics might change over time, maybe you want to do an academic textual analysis of how Jayce's ethnicity is approached by fic authors, maybe you're curious about how long Jayce/Victor fics tend to be or what other ships tend to show up in this search. These are all things that you can do when you're looking at a list of the actual fics.

And these are all things that you can't do with a score that does not count actual fics.

--

Since the numbers are scores and not counts, we can neither trust nor critique the info in other columns.

Multiple rounds of these stats have inspired wank over the fact that Hermione's ethnicity is listed as "ambiguous". Because the chart lists a score, all we can do is wank about canon. But if, instead of a score, if we're looking at the actual nine thousand Hermione/Draco fics posted in the timeframe in question, we can further filter that search and see that only 33 of them are tagged POC Hermione. And if we have the time and skills for it, we can do a textual analyses of those fics and see how she's protrayed when an ethnicity isn't tagged.

Last January I made a stink about Kirk/Spock being listed as a ship from the reboot movies, despite All Media Types and TOS being more common for K/S fics posted in 2024. I noted at the time that we have no way of knowing if the reboot movies are actually the top fandom tag in the fics in her dataset or not, but on further reflection I realised that her dataset contains no fics at all. New fics are not excluded from the set when old fics are deleted, because her data was never counting fics that exist in the first place. It's a point based score system that cannot be replicated or analysed, and that means absolutely nothing.

AO3 released its ship stats for 2025 where it also includes the race of the characters. There were people complaining that fandom is racist because there are so many white ships (someone even said that anime characters don't count) and in the same thread there were complaints that white people fetishize POC. Then there was a weird thing where they said that white people write all the Asian ships I guess forgetting all the Asians that are on AO3 writing fic.

Okay so what the fuck do they want us to do? Do they want us to write POC in ships or not?

--

Okay, can people stop saying that AO3 releases ship stats?

AO3 does no such thing.

Individual fans who scrape AO3 or look up totals from works searches release stats. Which stats they choose grab and how reliable those stats are vary. These are individual private projects that depend on the taste of the fan doing them.

--

As for what complainers want, they want you to stop writing and become the audience for their fic. Duh.

And they want to score internet points.

Block anyone who is annoying about Asian fandoms even if you aren't in Asian fandoms. It's a great litmus test for who is terminally irritating and who knows fucking nothing about how fandoms work yet wants to write pan-fandom meta anyway.

Randomist has not posted their real stats, but they likely will within a few days (theirs takes longer than centreoftheselights', because they actually count real fics that exist). Watch this space:

An Archive of Our Own, a project of the Organization for Transformative Works

For the lengthy breakdowns on why centreoftheselights's states are fake and meaningless, you can scope out my Bad Ship Stats tag.

The super short version is that she does not actually look at how many fics were posted in the given timeframe, but substracts last year's total fic count from this year's, which means that older works that were deleted, archive locked, archive unlocked or had their tags changed are erroneously counted in or subtracted from this year's count.

The older the fandom the more error this method introduces, which is why juggernauts like Destiel are always far lower than one would expect.

Also, the character races are just her best guess, and she is not in every single fandom on the list. Her guess is as good as mine or yours.

The Bad Ship Stats thing gets worse. There are bonus charts, some of them are breakdowns of top ships by canon's region of origin.

https://archiveofourown.org/works/58277833/chapters/149690638

Good Omens and The Witcher made the North America list and not the Europe list, presumably because an American production company for the TV shows overwrites European authors of the books the shows are based on, European cast and crew of the shows, and, in the Witcher's case, a European game studio.

Also for the Witcher, I feel like listing it as simply US American fantasy does a disservice to the Polish folklore and culture the books are rooted in.

The Witcher is even listed by its All Media Types tag, which includes the original Polish title. Centreoftheselights is literally out here saying that Wiedźmin is an American canon.

This does not surprise me. I've also seen many people point out issues with their race classification system (and their AO3 demographics survey last year was a hot mess). I know the gender/race/culture aspect of this "statistics" project is kinda their whole thing and why they're supposedly doing it in the first place, but if it were me I wouldn't touch any of that with a 10 foot pole. There's just too much nuance and ambiguity to be accurately judged by a single person, especially when you start getting into non-human characters and fantasy settings and especially considering that the person making the calls is not in most of the fandoms or a member of all races and cultures. It's just a recipe for disaster, even if their data collection methodology were solid to begin with.

Speaking of categorization inconsistencies, please allow me another quick side rant.

It annoys me that they claim that the race categories are based on audience perception (specifically North American perception, though this is unstated) rather than character self-identification, and yet when it comes to gender Aziraphale/Crowley is classified as "other" rather than M/M when we all know it's perceived by the general audience as M/M and wouldn't have anywhere near the same popularity if the actors were a different gender. Pick a lane! Either in-universe identification matters or it doesn't!

Yeah, it's a huge can of worms that I wouldn't want to open if I was tallying up the popularity contest winners. I glance at the comments for the year end stats on AO3 and its an even split between "Did Destiel just somehow get missed???" and "on what planet is my Blorbo the race you say they are?". I can only critique the listed races for fandoms I'm familiar with, and this year that's very few, but last year there were updated lists to correct the listed races for Genshin Impact and Trigun characters because she isn't familiar with those canons just plain guessed based on the anime skin tones, only for people who actually knew anything about those characters to correct her. (I can attest that her assumptions about Wolfwood were a bit bonkers, but for the Genshin blorbos I just have to take other people's word for it the way she has to. And that means assuming the folks in the comments are being objective and not pushing their own subjective headcanons or agendas).

In a list with so so many nonhuman characters and characters from fantasy settings that are not direct equivalents to real-world cultures and locations, it's just incoherent to list characters as White or Nonhuman based on skintone alone. And this year the post clearly states "N.H. = Non-Human Skin Tone"

There are interesting conversations to be had about things like what it actually means to say a character like Spock is white and how a largely American audience percieves a literally paper-white anime-style character from a Chinese game, but just going down the list with a handful of pantone swatches is not how to do that.

I'm also a bit cranky that she's using actor's ethnicity to determine a character's race when it's unstated in canon, but then only (one of several) production companies matters when determining a work's culture or location of origin. We could at least have a bit of consistencey here.

I've been squinting at her methodology since she completely ignored my feedback on the AO3 demographic survey prelims about how forcing all respondents to use the split attraction model will confuse some users and be wildly inaccurate for others, and everything else I've seen from her since has been at least as shoddy.

#the dismissive way she treats actual poc with concerns and criticism of how she categorizes race is also fucking crazy #i've seen at least 4 different latinos in the comments of this year's garbage data pile complain about how she's categorizing things #and her response is so fucking dog water each time #diría que es una pinche gringa pero ni eso es #she's a white european of all fuckin things #like god shut up and stop white knighting we don't want your bad stats #statistics #ao3

And another thing!

The tag listed in the "fandom" column is not necessarily the fandom the ship is associated with, but the top fandom within that ship tag over all time. It's inaccurate (I'm going to make that a separate post), but it's data that was scraped from the archive.

But the gender category is not the top category in the ship tag, it's based on Centreoftheselight's perception of the characters' genders. The top category for Astarion/You is overwhelmingly F/M (2043 out of a total 2589 fics for the entire tag) , but she lists that ship as Other to account for small minority the ambiguous, male, and nonbinary reader characters.

And Aziraphale/Crowley has M/M at the top with 57060 fics out of 70,224 total.

Even setting aside criticizm of injecting her subjective opinion about these ships and characters into what should be a factual report of the numbers, and criticism of the fact that some of the fandoms are just plain wrong, this means that the different columns of data within the same chart do not represent the same type of data. The "works gained" and "total works" columns list real (but misrepresented) numbers that exist, the "relationship" and "fandom" columns lists real AO3 tags that exist and actual data (the latter of which is again being misrepresented), but the "Type" column looks like real AO3 tags, but but is actually a list of the OP's subjective opinion, headcanon, or best guess, just like the "race".

When it comes to misrepresenting what the chart actually shows, the "type" column might be the worst offender.

owenhcrper

Milana Vayntrub as Olesya Ilyukhina Project Hail Mary (2026)

#olesya my beloved <33 #phm #olesya ilyukhina

fictionalauthorityii

Finding out AI companies are chewing through - literally, they’re pulping and burning them as they go - millions of books often including 1/1s and old texts that cannot be reprinted

Meanwhile there’s always some dipshit telling me AI is the future and we’ve got to just let companies erase human history and literature so they can train their abomination machines to pretend to think like people because a judge said it’s fine

wageronancap

odessa-edmundson

AI companies are buying and scanning printed books for training data, raising copyright and cultural preservation concerns as rare volumes a

#hey girl. what the fuck #can we stop actively fucking creating the Torment Nexus #anti ai

AO3 Ship Stats: Year In Bad Data

You may have seen this AO3 Year In Review.

It hasn’t crossed my tumblr dash but it sure is circulating on twitter with 3.5M views, 10K likes, 17K retweets and counting. Normally this would be great! I love data and charts and comparisons!

Except this data is GARBAGE and belongs in the TRASH.

I first noticed something fishy when I realized that Steve/Bucky – the 5th largest ship on AO3 by total fic count – wasn’t on this Top 100 list anywhere. I know Marvel’s popularity has fallen in recent years, but not that much. Especially considering some of the other ships that made it on the list. You mean to tell me a femslash HP ship (Mary MacDonald/Lily Potter) in which one half of the pairing was so minor I had to look up her name because she was only mentioned once in a single flashback scene beat fandom juggernaut Stucky? I call bullshit.

Now obviously jumping to conclusions based on gut instinct alone is horrible practice... but it is a good place to start. So let’s look at the actual numbers and discover why this entire dataset sits on a throne of lies.