H
HandsInTheDark
Guest
Generally I've decided that ratings don't mean much, and I've said so. But I recently posted a chapter in Non-Erotic, and watched the voting. As befits a later chapter, only readers into the concept are still around, so ratings were generally wonderful: 5,5,5,5,5,5,5,5,5,5,1,4,5,5.
Spot the outlier.
There's nothing unusual about this pattern. I've got a hater who shows up every few days to see if I've posted anything. It's a guaranteed one bomb. I've got a theory as to who he is and why, but that doesn't matter because there's squat I can do about it. Most of you probably know the feeling; it's just how Literotica works.
Normally it doesn't bug me. This time, though, I'm writing in a category where votes are uncommon, so it's going to take awhile for that 1 to get averaged away. It's dragged down the score to 4.64, low enough that people are likely passing it by. Which is the point of an early one bomb, of course. Sooner or later, Manu will get around to sweeping it, but by then the story is off the new list, so damage done.
So I'm back to thinking about ways to deal with this crap.
What I'd like is for Literotica to switch to computing a geometric mean instead of an average. It's not as cheap to compute, and requires a little special programming for large sample sizes because it gets into bigint territory fast. But it does what I think Lit should have been doing all along, showing what most people think about a story, and reducing the impact of outliers. That set of 14 votes yields a 4.84 as a geometric mean - which I think is much closer to what people are generally saying about it.
I think there are a lot of advantages. It would free Manu from bothering with sweeps, except perhaps during contests when ranking is extremely close. Onebomers would eventually figure out that they just don't have much impact, and maybe some would give it up. It also diminishes the effect of your best friend's undeserved 5. Contests would feel less rigged, though a large enough voting bloc, if those exist, will still prevail.
I'll send this on to Manu because he's assured me that Changes Are Coming This Year. I'll post it here so people can kick around the idea.
Spot the outlier.
There's nothing unusual about this pattern. I've got a hater who shows up every few days to see if I've posted anything. It's a guaranteed one bomb. I've got a theory as to who he is and why, but that doesn't matter because there's squat I can do about it. Most of you probably know the feeling; it's just how Literotica works.
Normally it doesn't bug me. This time, though, I'm writing in a category where votes are uncommon, so it's going to take awhile for that 1 to get averaged away. It's dragged down the score to 4.64, low enough that people are likely passing it by. Which is the point of an early one bomb, of course. Sooner or later, Manu will get around to sweeping it, but by then the story is off the new list, so damage done.
So I'm back to thinking about ways to deal with this crap.
What I'd like is for Literotica to switch to computing a geometric mean instead of an average. It's not as cheap to compute, and requires a little special programming for large sample sizes because it gets into bigint territory fast. But it does what I think Lit should have been doing all along, showing what most people think about a story, and reducing the impact of outliers. That set of 14 votes yields a 4.84 as a geometric mean - which I think is much closer to what people are generally saying about it.
I think there are a lot of advantages. It would free Manu from bothering with sweeps, except perhaps during contests when ranking is extremely close. Onebomers would eventually figure out that they just don't have much impact, and maybe some would give it up. It also diminishes the effect of your best friend's undeserved 5. Contests would feel less rigged, though a large enough voting bloc, if those exist, will still prevail.
I'll send this on to Manu because he's assured me that Changes Are Coming This Year. I'll post it here so people can kick around the idea.