How are "similar stories" selected?

[jehoram]

I know this question has been brought up before, with speculation on the algorithms used to generate the list... notably in this one:

https://forum.literotica.com/threads/whats-been-showing-up-lately-as-a-similar-story.1598224/

but I don't think there's ever been a definitive answer.

From what I can figure out, there are a few ways to do it:
1. Go by whether people who favorited your story have favored others in the same category
2. List stories closest to the key words you specified, or in the story description
3. Go by the highest ratings, period
4. All of the above, but weighting them in some arcane manner.

Has anybody figured out what's going on?

 

[OverconfidentSarcasm]

As far as I can tell, it's totally random.

I find stories in that list that share MAYBE two tags, but are in completely different categories, written by different authors, have vastly different ratings, and the system DEFINITELY isn't able to detect plot similarities, because many times they have absolutely nothing in common with the story they are supposedly similar to.

Though, and this could totally be just my imagination, it seems to have become better since they recently worked on that feature. Ever since it now shows the category the stories have been posted in, I find those lists to be actually somewhat accurate. Maybe they are now looking at user-created reading lists or something like "users who favorited this story also favorited...".

If the admins ever made a statement regarding that feature, I sure wasn't able to find it.

 

[EmilyMiller]

How are "similar stories" selected?​

I have no idea, but I had every single one of them being one of my own stories one time - so helpful

 

[Bazzle]

Well...
When I was helping out @Mericalovess with her stories the algorithm would put my stories in the suggested section.
So lit could work out we were working together and that our stories were similar?

 

[8letters]

The story that has appeared most often as a similar story since 8/30/2023 is the Mature story My Mom’s Best Friend by silkstockingslover.
Stats:

Tags:


The similar story count by category:


Erotic Horror? Really? Here are the eight Erotic Horror stories that have "My Mom's Best Friend" as a similar story:

 

[SimonDoom]

It's not random. It's crystal clear that there's an algorithm and it's not random. "Similar stories" tend to be the most popular stories that are somewhat similar to one's own story or that appear to have overlapping fans.

Long-term, it appears to be based on common favorites, i.e., it looks at the favorites that your story got, and it lists the stories that were most often favorited by the same people that favorited your story.

But this doesn't explain the INITIAL similar stories list, because until people start favoriting your story there's no way to generate a list of commonly favorited stories. So I assume the initial list is based upon common tags, and this algorithm gives way over time to an algorithm based on common favorites.

 

[Lovecraft_Lore]

I think it starts with common tags. I used some strange tags in a few different stories. Those appeared as similar stories when new stories with those tags appeared.

 

[8letters]

If anyone would like a list of their stories that have appeared since 8/30/23 as a similar story, start a conversation with me and include the URL to your list of submissions.

 

[OverconfidentSarcasm]

It's not random. It's crystal clear that there's an algorithm and it's not random. "Similar stories" tend to be the most popular stories that are somewhat similar to one's own story or that appear to have overlapping fans.

Long-term, it appears to be based on common favorites, i.e., it looks at the favorites that your story got, and it lists the stories that were most often favorited by the same people that favorited your story.

But this doesn't explain the INITIAL similar stories list, because until people start favoriting your story there's no way to generate a list of commonly favorited stories. So I assume the initial list is based upon common tags, and this algorithm gives way over time to an algorithm based on common favorites.

And how is this "crystal clear"? Because I honestly can not say where that confidence comes from.

When I want to check something like this, I find it easier to go with the extremes. They are simply easier to distinguish. So, I went to the I/T category and found a story of a mother who starts sleeping with her son's bully, before she joins said bully in tormenting her son until the little guy tries to kill himself. It should be simple enough to put all the "Mom+Son's Bully" stories into the similar list, right?

But there aren't any. All the "similar" stories are simply Mom/Son stories, even though the mother in the original doesn't even touch her son.

Now, maybe you want to argue that the story isn't old enough, or that 50 favorites aren't enough. So, let's look at another extreme in the category: A mother who likes to cuckold her son by sleeping with all of his friends, that has 200 favorites and is six years old. More than enough time and data to find all the other stories where mothers sleep with their son's friends. But, surprise, in the list at the end of this story, there's ONE entry in the similar list that fits.

So, I really don't see how these lists make sense.

 

[SimonDoom]

And how is this "crystal clear"? Because I honestly can not say where that confidence comes from.

When I want to check something like this, I find it easier to go with the extremes. They are simply easier to distinguish. So, I went to the I/T category and found a story of a mother who starts sleeping with her son's bully, before she joins said bully in tormenting her son until the little guy tries to kill himself. It should be simple enough to put all the "Mom+Son's Bully" stories into the similar list, right?

But there aren't any. All the "similar" stories are simply Mom/Son stories, even though the mother in the original doesn't even touch her son.

Now, maybe you want to argue that the story isn't old enough, or that 50 favorites aren't enough. So, let's look at another extreme in the category: A mother who likes to cuckold her son by sleeping with all of his friends, that has 200 favorites and is six years old. More than enough time and data to find all the other stories where mothers sleep with their son's friends. But, surprise, in the list at the end of this story, there's ONE entry in the similar list that fits.

So, I really don't see how these lists make sense.

You don't see the pattern because you are looking for the wrong pattern. Forget exact subject matter similarity. The key is common favorites.

My similar stories lists show pronounced patterns, but not necessarily tied to specific subject matter.

So, my similar stories lists tend to consist of:

1) my own stories -- that makes sense, because readers who favorte one of my stories are more likely to favorite another of my stories, EVEN IF those other stories are dissimilar in subject matter.

2) stories that are somewhat similar and that are unusually popular in terms of ranking high on favorites lists. So, for example, I have a mom and son on a seat together story. The similar stories lists is made up largely of the most-favorited stories with somewhat similar subject matter--not necessarily the MOST similar stories. This makes sense, probabilistically. People who found this story and happened to like it enough to favorite it are much more likely to have found and favorited other highly favorited stories with somewhat similar subject matter. But an exact match of subject matter is irrelevant. It's not a predictor.

The more favorites your story gets, the more you will see this is true, because with relatively few favorites the random factor is much higher. The results become more predictable as the number of favorites increases.

 

[ElectricBlue]

I think it starts with common tags. I used some strange tags in a few different stories. Those appeared as similar stories when new stories with those tags appeared.

Not sure about that. I just checked my latest chapter (a week old) against its "similar stories" list, and I didn't spot any similar tags. But what I did see was a bunch of common words in the story content itself - the name of characters, for example, or a similar description of a person carrying bags.

I reckon it's more than tags, it looks like something digging into the actual content.

Intriguingly, one "similar" story was by EclecticBlue, so there's a possible clue - similar author names? None of the others were close, though. It's probably spin a wheel - I don't pay much attention, doesn't the list change over time?

 

[MyBareTorso]

Mine are now very often my own stories... perhaps because I (now somewhat regretfully) include the phrase "Oz Beach Boy" in *all* my titles, and because I have a few tags (CFNM, Muscular man, Big cock...excuse me!) that, again, feature in all my stories...

 

[MillieDynamite]

We have similar stories suggested? Where, by whom, come on, I want to know.

 

[8letters]

Mine are now very often my own stories... perhaps because I (now somewhat regretfully) include the phrase "Oz Beach Boy" in *all* my titles, and because I have a few tags (CFNM, Muscular man, Big cock...excuse me!) that, again, feature in all my stories...

From what I've seen of the data, it doesn't matter what's in your title or what's in your tags. See my post above. Why you are seeing your own stories as Similar Stories is because the people who have favorited one of your "Oz Beach Boy" stories have favorited many of your "Oz Beach Boy" stories.

 

[lovecraft68]

One of mine always shows up.

Its gotten better because in the past I'd see stories nothing like mine appear, (even GM stories from time to time) now everything seems to be close enough.

I think it's another thing people fret over that isn't that important.

 

[ElectricBlue]

From what I've seen of the data, it doesn't matter what's in your title or what's in your tags. See my post above. Why you are seeing your own stories as Similar Stories is because the people who have favorited one of your "Oz Beach Boy" stories have favorited many of your "Oz Beach Boy" stories.

That doesn't explain the "similar story" links when something is first published though.

Favourites might factor in later on, for sure, but they can't on Day One, because no-one's read the story yet. There must be something else, or it's a random wheel spin.

 

[Voboy]

If it's based on "Favorites," then it's actually fairly useful IMO. I find many enjoyable stories by seeing what's been favorited by people who favorite me. As far as I can tell, that's a fairly good predictor of whether you're likely to enjoy reading a given story.

 

[Bamagan]

I wonder how much the behavior of profligate favoriters gums up the work of the proposed algorithm. By which I mean, the readers who have a list of hundreds of favorites spread across the gamut of what Literotica offers. Do you weight them more heavily because of the volume of their contributions, or do you treat them as too noisy to have much value? I'd lean toward the latter, but a lot of 'engagement' algorithms are designed with the mentality of the former.

 

[M_K_Babalon]

I just don't pay attention to them. It's one of the things here that make lil to no sense, no matter how you theorize them. I just accept that they're there and try not to think of the method of the madnes implimented. At worst, every theory is wrong, at best(if you can call it that), every theory is right. Either way is gonna leave more questions. And the two people that do know, I don't think Batman could get the answers.

 

[AchtungNight]

I will note the similar stories to what I read on occasion, but I’m more likely to look at what else the author has posted and read those stories if interested. It’s always great when authors continue their ongoing character arcs and plots.

 

[MyBareTorso]

From what I've seen of the data, it doesn't matter what's in your title or what's in your tags. See my post above. Why you are seeing your own stories as Similar Stories is because the people who have favorited one of your "Oz Beach Boy" stories have favorited many of your "Oz Beach Boy" stories.

All good! Whatever the cause, I'm selfishly happy it's happening like that!

 

[KeithD]

Has anybody figured out what's going on?

When I've checked, there hasn't seemed to be much correlation at all, so I now more or less ignore it.

 

[SimonDoom]

That doesn't explain the "similar story" links when something is first published though.

Favourites might factor in later on, for sure, but they can't on Day One, because no-one's read the story yet. There must be something else, or it's a random wheel spin.

I haven't done this, but the way to figure this out would be to watch a story's similar stories list immediately after publication and watch how it changes as it accumulates favorites.

When I look at the similar stories list for stories of mine that have been published for a while and have accumulated many favorites, I notice that without exception the stories on the list themselves have many favorites, but not necessarily any tag similarity. For instance, the number 1 story on the similar stories list for my story BTB Incorporated is An Unexpected Reaction, also a Loving Wives story, but one with no tags. It has over 2 million views and almost 3,000 favorites, however.

 

[Britva415]

.