8L Stats: Average rating by category and page length

8letters

Writing
Joined
May 27, 2013
Posts
2,160
I posted this in the Ratings thread and am reposting it for those people who didn't see it there.

Ratings are strongly influenced by category and page length
1699889908158.png
The first number is what percentage of all stories published in that category have that many pages. The second number is the average rating for that category and number of pages. This is for stories published from 8/30 to 10/13, 28 days after they were published. So 55% (39 out of 71) of the stories published in Anal from 8/30 to 10/13 were 1 page in length. Those one-page stories had an average rating of 4.28 on 28 days after they were published.

In my simplistic nerdish brain, I think that authors would be grateful for me providing this kind of information. You want to know how you are doing compared to average? Here you go! But when I think more on it, I can see this kind of information pissing off a lot of authors. "I have a 4.67 rating on my three-page Gay Male story, and you're telling me that's below average? Fuck off, asshole!"
 
The thing that surprises me most isn't the mean ratings as much as the percentage of relatively short stories. In Incest, for example, a category that sees many stories every single day, 68% of stories are under 3 pages, which I would not have guessed. The data certainly undercut any claim that Literotica readers only want short "stroke" stories. Anything over 2 pages is likely to have at least some development in it.
 
I posted this in the Ratings thread and am reposting it for those people who didn't see it there.

Ratings are strongly influenced by category and page length
View attachment 2288599
The first number is what percentage of all stories published in that category have that many pages. The second number is the average rating for that category and number of pages. This is for stories published from 8/30 to 10/13, 28 days after they were published. So 55% (39 out of 71) of the stories published in Anal from 8/30 to 10/13 were 1 page in length. Those one-page stories had an average rating of 4.28 on 28 days after they were published.

In my simplistic nerdish brain, I think that authors would be grateful for me providing this kind of information. You want to know how you are doing compared to average? Here you go! But when I think more on it, I can see this kind of information pissing off a lot of authors. "I have a 4.67 rating on my three-page Gay Male story, and you're telling me that's below average? Fuck off, asshole!"
Loving Wives scores need to be normalized to fit the Gaussian provided by the other categories.

Em
 
While I can appreciate the effort, especially the effort of finding out how many pages each story occupies, this is basically just raw data. Without the number of stories in each category, the percentage doesn't mean much. As far as the effect of the number of pages having an effect upon the rating, I seriously doubt the length of the story has anything to do with the rating unless the story is either very short or very long, though the data doesn't support even that conclusion and can't without a listing of how many votes were cast. A low rating for a story with 60 votes will have a much larger impact than the same low rating for a story with 500 votes.

Just looking at the data, it would appear that the genre has more to do with the ratings than the number of pages. For instance, LW seems to generate average ratings between 3.3 and 4 for all stories, while incest seems to generate ratings between 4.3 and 4.6. Most other genre fall into about the same range regardless of the number of pages.

The other problem is that averages don't reflect the impact of votes unless you know each vote and analyze that data to determine if the distribution is normal. There are a lot of ways to get to an average of 4.3 with 100 votes.
 
While I can appreciate the effort, especially the effort of finding out how many pages each story occupies, this is basically just raw data. Without the number of stories in each category, the percentage doesn't mean much. As far as the effect of the number of pages having an effect upon the rating, I seriously doubt the length of the story has anything to do with the rating unless the story is either very short or very long, though the data doesn't support even that conclusion and can't without a listing of how many votes were cast. A low rating for a story with 60 votes will have a much larger impact than the same low rating for a story with 500 votes.

Just looking at the data, it would appear that the genre has more to do with the ratings than the number of pages. For instance, LW seems to generate average ratings between 3.3 and 4 for all stories, while incest seems to generate ratings between 4.3 and 4.6. Most other genre fall into about the same range regardless of the number of pages.

The other problem is that averages don't reflect the impact of votes unless you know each vote and analyze that data to determine if the distribution is normal. There are a lot of ways to get to an average of 4.3 with 100 votes.
You’re right of course speaking statistically. But you can still draw some tentative preliminary conclusions. Like the LW scores being artificially low (unless we claim that a majority of LW stories are just poorer quality than elsewhere).

Em
 
You’re right of course speaking statistically. But you can still draw some tentative preliminary conclusions. Like the LW scores being artificially low (unless we claim that a majority of LW stories are just poorer quality than elsewhere).

Em
I think the LW stories are low because some readers go there just to vote low and comment on how the wife should have been drawn, quartered, and her head put on a pike in the town square.
 
The thing that surprises me most isn't the mean ratings as much as the percentage of relatively short stories. In Incest, for example, a category that sees many stories every single day, 68% of stories are under 3 pages, which I would not have guessed. The data certainly undercut any claim that Literotica readers only want short "stroke" stories. Anything over 2 pages is likely to have at least some development in it.
I think we had this discussion when I first posted this data way back in 2018. I was totally shocked then by how many short I/T stories there were. I think most one or two page stories aren't memorable, so I don't think about them when I thing about how long the average I/T story is.
 
I think the LW stories are low because some readers go there just to vote low and comment on how the wife should have been drawn, quartered, and her head put on a pike in the town square.
Of course. But the existence of such people (sometimes it’s the author’s head they want on a pike as well - especially “deviants” like me) could be adjusted for in the scoring. Just a simple formula to make them line up with elsewhere.

Em
 
Loving Wives scores need to be normalized to fit the Gaussian provided by the other categories.

Em
@NotWise shared with me a formula for estimating the percentages of Literotica individual votes. LW followed the same formula, but had a lower constant value. I wouldn't be surprised if each category follows the same general forumula but has a different constant.
 
@NotWise shared with me a formula for estimating the percentages of Literotica individual votes. LW followed the same formula, but had a lower constant value. I wouldn't be surprised if each category follows the same general forumula but has a different constant.
I wouldn’t either. But LW seems to be an outlier.

Em
 
Using @8letters' compilation from August, LW is actually pretty consistent with the normal pattern of voting. It just has a lower average. Even the occurrence of one star votes in LW is pretty consistent with the normal pattern. Looking just at the vote distribution and not at comments, I'd have a hard time saying why LW is regarded as so unusual.

I'm in my office where I don't have the graphs I showed 8letters. They were also posted in the Coffee Shop thread. If there's some interest, then I can show them here.
 
Currently, unlike on Amazon, we have no way to track actual pages read by a download. A download is opening the story, so when some leave and come back, they register a new download when they do. Also, some don't read past the first line or maybe close it if they look at the tags. If the writer opens the stories and then jumps to the comments (as opposed to opening the comments), they get a hit from their opening the story.
 
What I find most interesting is the way that scores pick up in relation to the number of pages. I guess it's because if you've stuck with a story to page 6, you must like it. There is probably also a tendency that even if a long story wasn't perfect, a reader may give credit for the amount of effort put in by the author.
 
Currently, unlike on Amazon, we have no way to track actual pages read by a download. A download is opening the story, so when some leave and come back, they register a new download when they do. Also, some don't read past the first line or maybe close it if they look at the tags. If the writer opens the stories and then jumps to the comments (as opposed to opening the comments), they get a hit from their opening the story.
But doesn't that mean that an unscrupulous author could repeatedly open and close their own story to inflate the read count?
 
I posted this in the Ratings thread and am reposting it for those people who didn't see it there.

Ratings are strongly influenced by category and page length
View attachment 2288599
The first number is what percentage of all stories published in that category have that many pages. The second number is the average rating for that category and number of pages. This is for stories published from 8/30 to 10/13, 28 days after they were published. So 55% (39 out of 71) of the stories published in Anal from 8/30 to 10/13 were 1 page in length. Those one-page stories had an average rating of 4.28 on 28 days after they were published.

In my simplistic nerdish brain, I think that authors would be grateful for me providing this kind of information. You want to know how you are doing compared to average? Here you go! But when I think more on it, I can see this kind of information pissing off a lot of authors. "I have a 4.67 rating on my three-page Gay Male story, and you're telling me that's below average? Fuck off, asshole!"
I see the biggest "caveat" to this data being the inability to distinguish any difference between a chapter versus a complete story and how that affect the scoring within the categories.

Take the Novels/Novellas category as an example...

At approximately 3,500 word per Lit page, a story would have to be a minimum of 5 pages long to be considered a novella (17,500 words [12 pages would be the minimum for a novel (40,000)]).

Since 78% of the "stories" in this category are less than 5 pages long, it would stand to reason that these are chapters or parts of the story.
 
Yeah, but that only drops their vote count, so it's a two-edged sword. But we don't have a read count. We have a download count. No one knows if there are any actual reads. I don't know what advantage a high read count gives you other than some list for the most downloads. Your front page listing in a category isn't based on reads.
But doesn't that mean that an unscrupulous author could repeatedly open and close their own story to inflate the read count?
 
Thank you for drawing that up, it validates my unmathmatical theory that a longer story will received a higher score because the reader has to truly hang on until the end. If you write a long bad story the readers will drop before they reach the vote.

However my unscientific brain would like to know how no stories can get an average score. In Group sex there were 0 stories of six and seven pages in length with fairly respectable scores. How do I submit nothing and get 4.90?
 
But doesn't that mean that an unscrupulous author could repeatedly open and close their own story to inflate the read count?
Yes they could. Anyone could.

I did a test once on one of my stories. I opened the story and refreshed the page 100 times. The view count went up by 100.

Views is simply page loads, it's an old web1.0 way of traffic counting. A more modern way would be to disregard repeated loads from the same ip address or using a timeout, or using a tracking cookie and/or login credentials to ignore repeated views.
 
I see the biggest "caveat" to this data being the inability to distinguish any difference between a chapter versus a complete story and how that affect the scoring within the categories.

Take the Novels/Novellas category as an example...

At approximately 3,500 word per Lit page, a story would have to be a minimum of 5 pages long to be considered a novella (17,500 words [12 pages would be the minimum for a novel (40,000)]).

Since 78% of the "stories" in this category are less than 5 pages long, it would stand to reason that these are chapters or parts of the story.
When I did this back in 2018, I shared the numbers for all stories and then said that was junk. I then shared the numbers for only stand-alone stories. Over the years, I've come to disagree with that view. SF&F is ~80% chapter postings. N/N is something similar. Does looking at only stand-alone stories in those categories really tell you anything?

Ideally, I'd do this analysis for all stories, and then a similar analysis for only-stand alone stories, and then for only first-chapters in a series, and then for only non-first-chapters in a series. But you know, I only have so many hours in a day.
 
When did you do that test? I've tried that as a test, and it didn't change by refreshing my screen. I picked an old story, so it wasn't getting a lot of downloads, and I did it fie times, and he didn't change at all.
Yes they could. Anyone could.

I did a test once on one of my stories. I opened the story and refreshed the page 100 times. The view count went up by 100.

Views is simply page loads, it's an old web1.0 way of traffic counting. A more modern way would be to disregard repeated loads from the same ip address or using a timeout, or using a tracking cookie and/or login credentials to ignore repeated views.
 
When did you do that test? I've tried that as a test, and it didn't change by refreshing my screen. I picked an old story, so it wasn't getting a lot of downloads, and I did it fie times, and he didn't change at all.
You have to wait up to 10 minutes to see the results. Views only updates about once every 10 minutes.
 
After the first opening today.

Screenshot 2023-11-13 142430.png

After 40 reloads

Screenshot 2023-11-13 142538.png

Only nine more views are logged at this point. Whether it is from my reloads or other's opening, I don't know. But I did 30 reloads to get nine more hits.
 

Attachments

  • Screenshot 2023-11-13 142036.png
    Screenshot 2023-11-13 142036.png
    10.2 KB · Views: 2
Back
Top