Being against evil doesn't make you good.
I've been a bit under the weather since Trotcon, but feeling better now.
This weekend I need to work on putting together some poni stuff for my niece (she's up here for her birthday), and my inability to do basic things makes that a bit difficult. So I may be delayed a few days on the Trotcon retro and the other long post I still really really want to share with y'all.
To head off any possible confusion, I've added a clarification to TEEE's story page and a note at the top of the chapter explaining that TEEE was not written using LLM generative AI (the story actually predates this technology by several years).
[Adult story embed hidden]
I remain alive, and as of June 13th am now the number of symmetries in a cube.
I will be at Trotcon.
Please contact me if you're there! (Or even if not, that's okay too.) 
I might do an author party. I'll announce it with another post. Signal boost would be useful.
Sorry for disappearing. Ironically the thing I wanna talk about is the thing that keeps me from being here at FF or getting anything done. 
Granted, the M-filter is turned on, but still...
Page generated in 0.025 seconds
Total duration
975 users online
379,896 hits today, 2,728,293 yesterday
My Little Pony: Friendship is Magic Fanfiction
Designed and coded by knighty & Xaquseg - © 2011-2024
Support us
SubStar
Chat!
Discord
Follow us
Twitter
MLP: Friendship is Magic® - © 2024 Hasbro Inc.®
Fimfiction is in no way affiliated with or endorsed by Hasbro Inc.®
No, I think I did that with Stick Up on Friday. Lower number of general webpage views with a higher frequency of that specific story hits makes for more dramatic moves.
4930806
I was mainly kidding.
The real measure of success will be if it trumps all the porn in the "real" feature box.
Well deserved. I am looking forward to reading more.
Hey, congrats!
If nothing else, it has a supreme hook in the description. :D
Congrats, Trick! I'll be reading this later today...to much work in the yard this weekend. Blergh!
4930807 Oh, all that takes is a second chapter where... um... No, better not. Cadence is the Princess of Love, not.. Ahem. Yes. Moving onwards.
Congrats!
But yeah, the feature box works based on views per time, compared to all other stories.
4930867
It's (among other things) views per time period, plus an age penalty. You can never beat a story in the feature box based on views per time period; you can only beat it when its age penalty becomes high enough. If your story gets a few views right after posting, and a story in the featured box is very old, you win. That's how many stories get featured.
The thing that is most-important to getting featured is how long the stories currently in the featured box have been there. It can only be called buggy, and could be easily fixed, and I've told mods repeatedly how to fix it, and they've never listened, seemingly because they don't understand the concept of mathematical modeling.
4941901 Honestly this model isn't all that bad. The whole point of the feature box is to feature stories for a while, and then feature other ones. That stale popular stories can get knocked out of the box by fresh, yet not-that-popular stories (that still have more views/time than the competition) is pretty much how it should be.
Not perfect, by any means, but very usable.
4942212
The crucial flaw is that the algorithm doesn't consider whether a story is in the featured box or not when counting views. Stories that are in the box therefore beat all the stories that aren't by a factor of at least 10. This guarantees that any story that can stay in the featured box for 15 minutes, will stay there for at least 2 days, during which there is no competition for that slot--the story already in it is going to stay in it, no matter how much more popular some new story might be when measured not in absolute views, but in predicted views conditional on being / not being in the featured box.
As there are only 7 stories in the mature box, and each lasts 2 or more days, there are 2 to 4 windows per day, maybe 2 hours each in length, during which a story can be featured. So--excepting authors with massive followings--only maybe 1 in 4 stories posted has even a theoretical chance at being featured.
This isn't what the algorithm is supposed to do. The big problem isn't that it's based on bad measures of quality, like view count. That's a problem, but not as big as the totally broken way that the featured box algorithm ignores the distinction between views received while a story is and is not in the featured box.
4942215
Ok, that statement is totally false, and I know from personal experience. I've had stories in the box for 15 minutes, an hour, or 12 hours, before getting kicked out by some other fic gaining popularity. Anyhow, I see your point, but, again, speaking from personal experience, I really don't think the severity of the problem is anywhere near the level you think it is, that's all. I think much of the issue you're describing is little more than the snowball effect of a story being in the box and getting more views than it would otherwise. Which, yes, maybe the algorithms should consider that a story in the box has an advantage over all the rest, but... idk, fine tuning that would be difficult.
Also I doubt the system is advanced enough that it makes a prediction of possible views... that just seems like an absurd, impractical method that wouldn't accomplish anything. Are you telling me you know that for a fact, or just based on your speculation?
4942269
Yes, that's the issue. But it's a big problem. I've watched the box a lot, and while it's not uncommon for stories to get on the "immature" featured box and then get kicked off, it's rare for a story to be on the "mature" box for half an hour and then get kicked off. I don't think I've seen it it happen in the whole 6 years I've been watching, though I can't swear to that. I have a large dataset from a one-week period when I had a program record what story was in the box once every 10 minutes, which I could check for a more definitive answer, but I know I've graphed it and didn't see any instances of that happening.
I don't think it does. I wrote, "no matter how much more popular some new story might be when measured not in absolute views, but in predicted views conditional on being / not being in the featured box", not meaning that the box did that, but that if you did that, you'd find that the stories that now stay in the box for days would instead get rotated in and out pretty often, and many more stories would get some time in the box, instead of it being like winning a lottery.
Predicting values is the basic task of statistical modeling. It isn't difficult or advanced. At its most basic, it's just writing Y = mX + b and finding good values for m and b.
4930824
That means you're going to read it, right?
4943849
Someday. :B