#The (Data) Science of League fanfic

Long story short the Hugo-award winning fanfiction website archiveofourown.org is pretty easy to scrape for fanfic metadata. Unlike fanfiction.net, it’s got a clean and tidy interface and is exceptionally well tagged.

This allows for some fun analytics. My goal was to learn as much as possible about the lore and fan preferences without reading a single line of text.

The first thing I wanted to find out was whether fanfic was on the rise or not. We know League of Legends isn’t as popular as a game as it was in its peak in 2015-2016, but what about the characters and its world?

Here’s the amount of fanfic on both sites sorted by month: https://i.ibb.co/7kpYJcF/plot-new-works-count-by-month-sites.png

As we can see fanfic is alive and well, with a small increase in the last year or so, owed to the fact that Narrative has undergone a mass expansion, but also a few factors that will become obvious later.

Also, as we can see, fanfiction.net is dying. This is because it’s slow and poorly kept, has barely any metadata, doesn’t support tags and… because it doesn’t allow certain kinds of fiction. For these reasons, and despite the fact it contains many literary classics I will not base much of my analysis on it.

Alright, so we have our awesome metadata from Archive of our Own. What do we do with it?

#Characters with most fanfiction

I wanted to find out which characters (not just champions, Boram Darkwill and Yone can feature in stories) are most popular among fan writers. The results show a pretty clear picture:

https://i.ibb.co/TB1vqDG/plot-works-count-by-character.png

Aaaaaand it’s Lux, leading by quite the margin. The top ten being: Lux, Ahri, Akali, Ezreal, Zed, Evelynn, Katarina, Darius, Kayn, Caitlyn.

These are widely regarded as the Main Characters of LoL, who get the majority of content in both visual and narrative, they have resonance and broad appeal and it’s hardly surprising they feature in so much fiction. I want to pay close attention to the surprising oddball in 11-15, however:

11. Jhin
12. Shen
13. Riven ##14) Talon
14. Garen

##Characters with most fanfiction where they are the main character.

A lot of fanfics focus on a handful of characters, while mentioning many others, which doesn't give a very clear picture of writer interest. My criterion was simple here: The two characters with most mentions in a first chapter of the fic are main characters. It’s not perfect, but it works ~95% of the time, especially since many fics are oneshots. https://i.ibb.co/4pP2rdN/plot-works-count-by-main-character.png

Writer dedication

We can now compile a stacked bar chart from the above two. I present the proportion of fanfics in which a character is among the two main characters:

https://i.ibb.co/pQHN3Wj/plot-main-character-ratio.png

I think this plot measures more than popularity, it measures dedication/interest. It seems dominated by character who fill unpopulated niches (Lamb, Graves), and some mad devotion magnets like Diana and Kayn.

Speaking of writer devotion, I also took a Rioter’s suggestion to look at average works per author per character:

https://i.ibb.co/wsrFDmd/plot-dedication-character.png

Tag analytics

Did I mention ArchiveOfOurOwn has amazing tagging? It really allows you to categorize fanfic well. Here’s fanfic split by rating (maturity): https://i.ibb.co/9nhMB7D/plot-works-count-by-rating.png

From this we can see what we’ve always known, namely that fanfic is thirsty work. Explicit is the second biggest category and Mature isn’t that much smaller than General.

I wanted to know how these ratios have changed through the years, so I made this: https://i.ibb.co/nRYbxwW/plot-new-works-count-by-month.png

From this we can see the ratios remain more or less stable, with thirsty fic being somewhat on the rise?

We can also do this per character: https://i.ibb.co/QmCRJxy/plot-chars-rating.png

Anyway,

Tags

As we mentioned before, fanfic is thirsty work. For these exact reasons the next picture contains some questionable tags. Before you brand writers as thirsty animals, however, note that extreme acts/themes are more likely to get tagged than mild ones to make sure they reach the correct audience. I will attach it here, open at your own risk - it yields curious results.

Category

You may notice Yaoi/Yuri/Het are low on the above chart. This is because they aren’t classified as tags but rather categories on ao3, sort of a tag for ship type. So, https://i.ibb.co/8mvxSGk/plot-works-count-by-category.png Notice the proportion of fics that don’t contain a ship (gen or []). It is quite tiny.

Ship popularity

How could I forget: https://i.ibb.co/F6s7bx8/plot-works-count-by-ship.png

Character-specific plots

The real fun part of Ao3’s extensive tagging is the amount of info it allows you to obtain for individual characters.

For example, here’s a chart of which characters are likely to appear in a fic featuring Vladimir: https://i.ibb.co/Vx10BcK/plot-character-appear-Vladimir.png

Pretty accurate, right? More importantly, we can plot who Swain gets shipped with: https://i.ibb.co/5TZGQSG/plot-character-ships-Swain.png

And here's the tags on Teemo fanfic: https://i.ibb.co/dGKn1rv/plot-character-tags-Teemo.png

We can pretty much infer most of the key points of the lore canon/fanon without having read a single line of lore (or fanfic), and that, in a way, is beautiful.

The connections don't end here. We can attempt to plot a full network of character relations. Make sure to zoom in for this one, it's some 8000 x 8000 pixels across.

https://i.ibb.co/yfYh89R/plot-char-network.png

Messy, isn’t it? We can spot nodes (Lux ahem) that connect to many others. This is because many characters feature as side/supporting characters in stories containing Lux, or other popular kids.

Even so, we can see cliques form, corresponding to lore groups/regions. Notice for example Azir, Nasus, Renekton and Xerath hanging out together, or Orianna, Viktor and Blitzcrank. Or the yordle champs.

If we want to focus on cliques/region groups, we can minimise the number of edges:

https://i.ibb.co/28WVdcp/plot-char-network-small.png

These make almost perfect sense. The exceptions are in cases where skin lines become more popular than canon verse (KDA girls bundled together) or where lore doesn't manifest in the game too well (Not much reason for Elise, Zyra and Evelynn to be together). That said another way to interpret this is in a "if you like character X then you might like character Y", which is why Karthus and Vladimir are connected.

Character-specific bar plots

Another thing about characters: Most of the bar plots above can be repeated on a character scale. This for example allows us to see trends in writing for specific characters. These won’t always tell us relevant info, but are good for evaluating how major events have influenced a character, e.g. the way Swain’s rework resulted in a massive boost in fanfic: https://i.ibb.co/h19Mh7b/plot-new-works-count-by-month-Swain.png

Or the way KDA did more for Evelynn than her rework: https://i.ibb.co/Z6BwNrb/plot-new-works-count-by-month-Evelynn.png

Finally, I had a lot of fun making these in my quest to become better at Pandas. If you come up with some fun metric you want explored and visualised, let me know!

EDIT: Note these observations are only relevant for the fanfic until 02/02/2020. Nothing's stopping some mad lad from waking up tomorrow, writing 100 000 words of insanely popular Malphite fanfic, and turning my data upside down.