A Look at Fimfic's All-Time Top 100 · 2:02am May 26th, 2016
So the other day I was breezing through Fimfic's top 100 most-liked fics (which you can access here) and I noticed a few trends. Hmmm, my mind wondered. What would happen if I recorded those trends? What would I learn about Fimfiction? Or about what people seem to upvote the most, in large numbers? And so, piqued, I decided to browse the first 100 fics of Fimfic all-time top list ... and see what sort of data I could pull.
Ready? There's some interesting things here.
First, length. Out of 100 fics, I was shocked to see that 71 of them were less than 15,000 words. Looking back I should have done 1-5k, 5-10k, and 10-15k all as separate categories, but for the purposes of me not wanting to go back and recount everything, we're going to go with the one category. Although about half of them (roughly), IIRC, were less than 5k words.
Honestly, this shocked me pretty good. Fimfic, it would appear, overwhelmingly prefers short, fics ... and the shorter, the better. However, long fics do have a place—the second highest portion of the top 100 goes to stories above 50k words (or novel length), with 12 entries. Not bad ... though still woefully small in scale compared to the juggernaut that is the short story. The novellete, 15-30k, comes in right behind it at 11 entries. And novella length, 30-50k, barely weighs in with 6 entries total.
Right, so just to clarify, of the top 100 most loved fics on Fimfiction, 71 of them are short stories. That's not even a close battle. Everything else trails so far behind they barely show up. I knew short fiction was popular, but I severely underestimated the margin by which it was winning that battle.
What to make of this ... that's up to you, of course. I just dug some data. Now, onto genre tags. And, for the record here, I ignored site-specific tags like AU, Crossover, Human, and Random, instead noting more conventional, widespread genres. For each story I simply tallied the genres they listed. And ... here's what I got.
Slice of Life is a juggernaut just shy of the short story, but clearly tied into it. 68 of the 100 stories I looked at claimed the SoL tag. Again, a whopping majority. Draw your own conclusions.
Not too far behind it (and often sharing a spot with SoL) was Comedy, with 39 out of 100 stories tagging themselves as such. Adventure (my personal favorite) came in 3rd of all places with only 12 tags. It also tied for 3rd with Sad.
Romance had 10 tags. Drama 5. Dark 4. So yeah, SoL had a 5 to 1 lead over the 3rd place tie.
Anyway, a fun little exercise. Certainly it says a bit about what Fimfics majority audience votes for; short SoL stories reign supreme. Makes for some interesting possibilities of discussion on Fimfic reading habits, too.
Anyway, just something interesting I threw together. What do you think?
I think it's less that people prefer short stories so much as it's harder for a long story to maintain good like/dislike ratios on every update. The people that press like and the people that press dislike are not the same people, and a higher percentage per capita of the latter category will show up if you write long stories.
Out of 90,173 fics in my database, 71941 of them (.78) are less than 15,000 words, 18232 are >= 15K words, and 5200 are >= 50,000 words. So there is no bias for stories under or above 15,000 words, but among stories over 15K words, there is a preference for those over 50K. Sample size in favorite stories is small, however, and the bias may not be in reader preference, but in writer preference. Authors are more likely to continue stories with high ratings.
Story tag counts:
An|3053
Tr|11359
Cr|14090
Ra|15820
Sa|18115
Hu|20059
Al|21888
Da|25279
Co|26844
Ro|28493
Sl|28775
Ad|30068
(This dataset is from before the new tags.)
The thing is, when you're looking at fimfic top-ranked stories, you're not looking at people's favorite stories. If you ask people for their favorite stories, they'll often say "My Little Dashie", "Eternal", "Past Sins", or "Fallout: Equestria", none of which are anywhere near the top 100. The top-ranked stories are the least-objectionable stories on fimfiction. These tend to be "Slice-of-life".
3972896
You're not incorrect, but that would boil down to a preference for short stories, I think.
Metrics are fun to play around with. I know I like comparing the numbers I get from reviewers on my big master review list.
That people have short attention spans and only want to read happy stories where nothing bad ever happens?
There's another part as well. This site is inundated in short stories. I really much prefer the longer stories. If a short story is part of a greater setting (other than just MLP, like sidestories on a larger series of stories), it's one thing, but I just feel that you need a longer story to give any OC depth, or to truly explore something. But at the same time, those short stories so far outnumber other stories that there was little chance of them not being the majority of the top 100, simply based on quantity.
That's part of why I tend to put a minimum word count of 100k when I'm doing a search for something new to read, unless it was recommended by a friend.
Chinchillax is big on metrics and data sheets too.
I noticed a whole lot of the real "big follower" people just have an absolutely huge number of short fics about a bunch of concepts or short jokes. If I had a "guide to how to gain tons of followers on Fimfic" I would put "write a ton of oneshots" up at the top.
Then probably something about following fads and writing shippy stuff.
But that's popularity and not U/D ratios, I suppose. For example, shippy stuff apparently will not get you into the top 100 too easily, according to the numbers here.
Obviously it's easier to read short stories, and it's easier to write short stories... but I also think the power of the Feature Box helps shorter stories succeed more, even with the bottom three slots reserved for updates.
Would chalk it up to short stories that do not require too much in terms of character development. Just cause a slice of feel-good, and you get upvoted, regardless of the quality. See early stories such as My Little Dashie as examples. Read it, and never understood why so many people loved it.
I guess a lot depends on how you measure "top 100".
I have no idea how Fimfiction calculates its ratings, but I'll do the same statistics using my own local archive and the method that I use locally to rank stories (coincidentally, Fimfarchive just came out with a fresh new version of the archive, so you can all try it out too :)). I calculate the ratio as:
likes/(likes+dislikes) - 1.9599/(2*(likes+dislikes)^0.5)
This gives me a 95% lower bound confidence on the mean - I can say that there is a 95% chance that the "true" like ratio is at least what this formula produces. As a result of this formula stories that have lots of votes rank higher than ones with fewer - a story with 100 upvotes and 10 downvotes (0.81566) will rank lower than a story with 1000 upvotes and 150 downvotes (0.84067). Also, dates don't factor into it directly, though older stories will have had more opportunity to accumulate votes so they'll be somewhat overrepresented.
So, sorting by that value, the top 100 , I get this breakdown:
< 15k words - 50 stories
15-30k - 11 stories
30-50k - 11 stories
50k+ - 28 stories
Genre breakdown: 30 Adventure, 10 AU, 51 Comedy, 3 Crossover, 8 Dark, 2 Drama, 10 Human, 11 Random, 24 Romance, 10 Sad, 1 Sci-Fi, 42 Slice-of-Life. I imagine that tag count is skewed by the fact that some of those tags are newer than others.
Only 4 are marked Gore, and 4 are marked Sex. 53 are marked "Everyone", 47 are marked "Teen", and *none* are marked "Mature". 85 are Completed, 13 are In-Progress, 2 are On Hiatus (though 4 of the in-progress ones haven't been updated in over a year so I wouldn't hold my breath on those).
When I sort by year last updated I get:
2012 - 47
2013 - 24
2014 - 15
2015 - 8
2016 - 6
Showing the date skew due to how I weighted the statistics.
Anyone want to trade further ideas on how to do statistical analysis of this dataset?
Fimfic's ratings system is really, really fiddly. It's very sensitive to downvotes. Take for example An Old Coot, currently the eight highest story on the list. It has 10,516 views, 1,618 upvotes, 212 comments... and only 20 downvotes. Guess what the ranking changes to if I change my upvote on the story to a downvote? 69th place. Give a try yourself; just pick a story, count what position it's at on the list, and then see how that changes when you add your single downvote. (And then fix it by changing it to an upvote so you don't get bad karma.) At a guess, most of the stories at that top 100 ranking could be taken off by less than a dozen downvotes.
Now, if there was a filter for the most upvoted stories, that might be useful. As is, though, however the algorithms behind the scenes are calculating the ratings, they just aren't really helpful or fair.
3973151
Well, I can very easily sort my local archive purely by number of "likes", so let's see how the stats play out that way. :)
< 15k words - 23 stories
15-30k - 4 stories
30-50k - 12 stories
50k+ - 61 stories
This is significantly different! Looks like big stories get more upvotes, in absolute terms. I guess that makes sense - big stories have more chapters and are published over a longer time period, providing more opportunities for people to see them and register an opinion. These top 100 still all have a high *ratio*, mind you - there are no weird cases with 1000 upvotes and 1000 downvotes.
Genre breakdown: 34 Adventure, 17 AU, 58 Comedy, 9 Crossover, 19 Dark, 5 Drama, 28 Human, 9 Random, 33 Romance, 12 Sad, 2 Sc-Fi, 38 Slice of Life, 2 Tragedy. This is actually very similar to the results of my statistical-ratio-based top 100.
12 are marked Gore and 13 are marked Sex. That's rather higher than the ratio-based top 100. Maybe gory and cloppy stories draw both upvotes and downvotes in larger quantities? That would push the ratio down while raising them in absolute upvote terms.
34 Everyone, 57 Teen, 9 Mature. Similar to the sex and gore thing, this sample has a larger amount of "adult" material I guess.
More variety in the status tags here. 73 Completed, 23 In-Progress (9 of which haven't been updated in over a year), 3 On Hiatus, 1 Cancelled. Assuming the increased number of in-progress works is significant, one possible explanation that comes to mind is that stories get more downvotes while they're in-progress but get a flurry of upvoting when they are completed, which drives the like/dislike ratios for completed stories higher. Just a hypothesis, mind you.
One of the Completed ones in this top 100 sample has actually been deleted off of Fimfiction, not sure how to categorize that.
By year last updated:
2011 - 1
2012 - 36
2013 - 24
2014 - 16
2015 - 13
2016 - 10
Interesting, there's more of an even spread on the dates. I wasn't expecting that, I expected a similarly strong bias toward earlier works since they have more time to accumulate upvotes.
3972927
I'd add that it may also have to be that a person's favorite story may or may not be representative of the group as a whole—otherwise we'd see a lot more favorites of major reviewers and writers being the most loved stories on the site (and yes, this does happen, but there are plenty that are not as well). People tend to be very specific with a favorite, while like you said, something that 9 out of 10 people think "Oh, that was nice, I'll give that an upvote" about will actually do better than the story that gets 6 out of 10 because it's a genre that appeals to only 6 of those 10 and the other 4 never look.
There's also the other angle to consider that many of fimfics readers openly admit that they're not readers, ie, they only read certain fanfics and don't familiarize themselves with conventional reading at all (which is a real shame). But it means they're going to base their upvotes and downvotes based around what caters to that interest, and ... well, while that's definitely not the whole reason fimfic has interesting numbers, it certainly does explain a lot of the "favorites" versus what's actually been highly rated by the community.
As far as the numbers of stories printed on the site, I'd still suggest that a lot of that has to do with the audience, both reading and writing. How many writers on Fimfic, percentage-wise, could actually write a 50,000 word story successfully, or believe they can? Or even want to? But pumping out 2000 words on Twilight's reaction to ... spam, for instance, can be the work of an afternoon and bring instant internet adoration from a crowd of a like-mindset, one that isn't willing to commit to reading a 50k+ word work, but will commit ten minutes for a quick, easy to digest morsel.
I'm spitballing and tossing out random thoughts here, but then again I was doing that when I grabbed the numbers in the first place. Either way, I think it's interesting to look at compared to, say, the book market.
3972941
Yeah. They can be bent in just about any direction once statistics come into play, but the flat numbers are just interesting to look at and wonder about.
3972956
I definitely agree with the short attention span thing (and it's an issue for all forms of literature, not just here), but I'd alter the nothing bad happens a little. I don't think they read for the nothing bad but rather for the quick burst of catharsis." Sort of like a quick hit of sugar. We know we can go eat a real meal and get a much longer-lasting, healthier buzz, but we go for the candy bar because of the sweetness and short but immediate burst. Perhaps a lot of readers, even though they'd be better served by a longer work, will just read the 1,000 word one since it'll be fast an easy and they'll get a taste of what they want?
3973014
Short stories are quick and easy to write, which definitely helps. See the above comment on catharsis candy bars.
3973071
Yeah. That reminds me, I should see what he's been up to lately ...
3973072
I hadn't thought about that. Now I want to see some actual experiments performed on that, because I wonder if you may be right about the feature box accidentally favoring those short works due to the way it's set up.
3973077
Yup. Catharsis candy bar seems to be a common trend here.
3973080
Interesting what the differences are, since you're not using Fimfic's weighting (which as pointed out by another comment, seem to heavily weight downvotes to a high degree) ... and I'll admit, I plugged some of my own fics into that method you gave just to see what number they gave off, though I'm not certain what order of operations you used. The results you pulled up are interesting, though, and it makes me wonder just how different Fimfics is that the results are so different.
Interesting data.
3973151
Yes, they really are. A single downvote can move you several hundred or even a thousand spaces at a go, which seems unusual.
3973384
like peaches but Hawaiian. Maybe I'll write that.
Also, as far as short stories and the future box ago, I be leave there's a multiplier bonus as stories get longer, up to like 5000 words or something. That's part of why it's so hard to get short 1000 word stories in versus like the 6000 and 7000 word ones.
3973384 My method isn't particularly rigorously thought out, I just didn't want to use a straight "likes/(likes+dislikes)" ratio because it seemed silly that a story with 10 upvotes and 0 downvotes should be considered "better" than one with 1000 upvotes and 1 downvote. So I hacked together something that seemed reasonable at the time.
I just spent a bit of time trying to re-derive my method so I could explain it better but I do this probability stuff so infrequently that it's taking me a while. I have to re-learn stats each time I use it. σ = 1.9599 is the standard deviation for a 0.975 probability, and σ/√n is the standard error of the sample mean of a population, but darned if I can remember why I'm multiplying √n by 2 to get a 0.95 confidence. I might be wrong, I never really expected to show these numbers in public. :)
If you're curious, here are your numbers as of right now:
Why Me? - 0.94742
The Dusk Guard: Rise - 0.94078
Carry On - 0.93676
Old Habits - 0.93628
The Definition of Strength - 0.93323
Emoticon - 0.92923
The Dusk Guard: Beyond the Borderlands - 0.91983
Trust - 0.91896
Remembrance - 0.91219
Hearth's Warming Cookies - 0.91038
The highest-ranked story on Fimfiction, by this method, is Sunny Skies All Day Long with a score of 0.97344. An oldie but a goodie. :)
Anyway, I have no idea what would really be a "fair" way to rank stories. Not really fond of factoring page views in since the "motivation" behind a page view is hard to guess. Upvotes and downvotes are subjective too, just look at how certain classes of fanfic get strong upvote or downvote signals irrespective of actual content or quality (FoE, Conversion Bureau, etc).
Because I couldn't help myself, I did this for the top 100 most reviewed stories on my list. Which is a much smaller pool mind you, but any excuse to compare the top stories on the site as a whole and my own numbers. Even if it means actually gathering the data since I don't keep track of wordcount or tags on the list.
Turns out about the same for length of stories. A few less of the really short stuff, but exactly the same number of novel-length and up. The genres are a bit more evenly distributed though.
Length of stories:
66 are less than 15k words
17 more than 50k
16 between the two
1 exactly 15k words
Now, these overlap but here are the genre numbers:
51 slice of life
27 comedy
27 adventure
18 romance
19 sad
17 dark
10 alternate universe
10 drama
6 Gore
4 Mystery
3 Tragedy
Funny because I love a good, long story. Some of my favorite fics or series of fics on this site are over the million-word mark. Things you can dive into and lose yourself in for two weeks.
3973072
Shippy stuff will get you lots of reads and followers, but it won't get you onto the top-rated page because people have opinions about ships, and whatever ship you pick, somebody will down vote your story without reading it just for having that ship.
To get into the top 100 you need about 100 upvotes per downvote, which means your story must be free of anything that might upset or irritate anyone. It must not have anything challenging in it. It must have a sweet or melancholy ending that doesn't question any conventional wisdom. Some of these stories are quite good, but they are almost all either funny or syrupy, and not at all a collection of fimfiction's best stories. (I don't know that it's meaningful to believe in a collection of best stories.)
3973420 You're probably using the same approach knighty does, which is to give a story a ranking that it can be assigned with 95% confidence. This is a bad way to do it, because it gives a huge advantage to stories with lots of views, which are exactly the wrong stories to give an advantage to. IIRC, the correct way to compute the expected true ranking is (likes + 1) / (likes + dislikes + 2). You can use Laplace's rule of succession to derive it, but the deriviation is a nightmare. This, however, while correct for an individual story, results in populating the top-stories list entirely with stories with very few views, because they have higher variance in their scores. I don't think this is a problem, though, because it will self-correct as people read stories off the top-ranked list and add more ratings to them.
3973405
Yes, there is. You can see its effect by comparing ranking on the Popular Stories list to ranking on the featured box. At the moment, my story "The Gathering" is ranked #2 in popularity, but is only in featured box slot #7 because it's 1200 words.
3973713 That wouldn't work for my purposes, though, because I'm sorting a local archive that nobody else has access to. Nobody is going to go read a story with few likes and dislikes just because it's high up on a list that's sitting on my computer at home. :)
As far as Fimfiction is concerned I agree - there should be more mechanisms to draw attention to things that haven't had much. Once long ago I even proposed that the featured box should have one slot that rotates through old stories every few hours essentially at random, as a way to potentially dig up and expose old gems that were missed back when they were first posted out of happenstance. Maybe basing that off of a ranking like this would be a good way to more intelligently pick stories to expose than simple randomness.
This may be a bit biased because -- especially in the olden days of Fimfic -- the SoL tag was often used more like an 'other' category or a catch-all than for what it's really meant to be.
Now that we've got some better tags, I'd expect that trend to decrease, but most of the highest-rated fics are going to be relatively old, because building up a rating like that takes time.
(I wonder if my good old Unicorn Horns Are Made Of Candy is still up there... *Checks* Damn, it's in the 800's now. Somebody must have downvoted it. At one point, that story was ranked 37th on Fimfic. But a position at the top can be very precarious... just a few downvotes and it's gone forever.)
To add something building off the idea that downvotes affect a story more than upvotes: downvotes are also, I would think, easier to get. There's quite a few folks out there who will downvote a story just based on the description alone, sometimes even just the tags. On the other hand, I don't know of many folks who will upvote a story without at least trying to read it. That alone would provide a fairly obvious boost to short, SoL stories since they won't generally have anything dark or shippy going on (two tags very likely to get you some automatic downvotes).
tbh it makes me wonder about the wisdom of structuring a voting/rating structure such that it places "avoiding downvotes" as the top priority of any author looking to make a name for themselves.