Can someone explain how thecrating story rating system works?

[onehitwanda]

I feel like we need a "Statistics 101" thread

 

[alohadave]

It'd be possible to adjust scores to a more centered distribution, but it'd be a lot less straightforward than an average. It'd make it much harder for people to understand movements in their scores; for instance, one implication of "grading on a curve" is that your score can be changed by votes on other stories even without any new votes on yours.

The scoring at SOL is pretty much impenetrable for this reason.

 

[SimonDoom]

Lots of them! If you look at something like Uber ratings, or most restaurant reviews, scores tend towards the high end of the scale. Giving an Uber driver anything less than a perfect 5 is practically a declaration of war.

It'd be possible to adjust scores to a more centered distribution, but it'd be a lot less straightforward than an average. It'd make it much harder for people to understand movements in their scores; for instance, one implication of "grading on a curve" is that your score can be changed by votes on other stories even without any new votes on yours.

I think the only way you could do this here would be to convert the raw score to a percentile score, or a 1-5 score where 3 is the median. A percentile makes more sense to me because people generally understand what that is, and it wouldn't be confused with the raw score. It would also provide a fairer basis for awarding red Hs. You could designate, say, the 75 percentile in a category as the minimum needed to get a red H.

I suspect none of this is likel and we're just blowing into the wind.

 

[SimonDoom]

I suspect most people would be pissed off if the score accurately and clearly represented where their score stood relative to others'.

 

[pink_silk_glove]

I suspect most people would be pissed off if the score accurately and clearly represented where their score stood relative to others'.

Ya think?

 

[Bamagan]

I think the only way you could do this here would be to convert the raw score to a percentile score, or a 1-5 score where 3 is the median. A percentile makes more sense to me because people generally understand what that is, and it wouldn't be confused with the raw score. It would also provide a fairer basis for awarding red Hs. You could designate, say, the 75 percentile in a category as the minimum needed to get a red H.

I suspect none of this is likel and we're just blowing into the wind.

Blowing into the Wind would be a great name for a story about oral sex on a windjammer, though!

 

[SimonDoom]

Blowing into the Wind would be a great name for a story about oral sex on a windjammer, though!

Or a Kansas song.

 

[Bamagan]

Or a Kansas song.

All we are is Dustin DeWendt. Ah aaah, la de da da dah.

 

[TarnishedPenny]

One point I always make, with all possible respect, is that it’s unwise to complain too much about the one-bombers. They’re a fact of life, like ants at a picnic. Equally important is the probability that they’re looking to cause annoyance. Like with a toddler shouting potty words, attention is what the trolls thrive on. The more one complains, the more they are rewarded. The sweeps deal with most of them anyway.

 

[Bramblethorn]

I think the only way you could do this here would be to convert the raw score to a percentile score, or a 1-5 score where 3 is the median. A percentile makes more sense to me because people generally understand what that is, and it wouldn't be confused with the raw score. It would also provide a fairer basis for awarding red Hs. You could designate, say, the 75 percentile in a category as the minimum needed to get a red H.

Yeah, "grading on a curve" is basically about mapping to a percentile and then, optionally, mapping the percentile to whatever distribution you want to end up with. But it requires making a lot of decisions about how to implement it.

If we just take the current Literotica scores and calculate a percentile against every other story on the site, ignoring things like category, then we end up reproducing a lot of the biases of the current system - e.g. LW stories will end up with low percentiles relative to HEA romances.

If we calculate percentiles separately for each category, and then use them to compare between categories, then we're assuming that the median LW story is exactly as good as the median story in every other category. That might be better than the current biases but it could be overcorrecting - it's possible that some categories genuinely do attract a better standard of writing than others (for whatever definition of "better" one chooses).

It's possible to finesse that by looking at authors who post across multiple categories: if authors who typically score about the 80th percentile in LW are scoring around the 90th percentile in SFF, say, that might help gauge a difference in the typical quality between those two categories.

Give me time and some stats textbooks and I could come up with a hideously over-engineered system that addresses these issues along with a few others (e.g. lower reliability of stories with fewer votes), and it would gladden my stats-nerd heart, but people would hate it because it wouldn't be transparent.

Here, we were able to look at the OP's numbers and tell them "you went from 728/158 to 729/159, that's why your score did what it did". That's not going to be possible with a percentile-based score because it depends on every other story in the running.

I love designing technical solutions but in the end, a technical "solution" that's not built around the human element isn't a solution, it's just a hobby for nerds.

I suspect none of this is likel and we're just blowing into the wind.

Never stopped us before...

 

[Bramblethorn]

I feel like we need a "Statistics 101" thread

Well hello.

 

[djrip]

Hmm, I probably need a new analogy since young 'uns these days will never have seen static snow on a TV.)

I used to stare at that stuff for hours... Well, minutes anyway.

 

[TarnishedPenny]

I used to stare at that stuff for hours... Well, minutes anyway.

And test patterns! Remember those? Duleigh will remember the Channel 4 Buffalo.

 

[MyBareTorso]

Thanks to everyone for their candid, and probably spot on explanations. Its just frustrating to experience...

There is certainly much that is frustrating and confusing about Literotica, especially with regards to scores etc. I'm with you on that. But look, some might see it as trite, but my response is to try and focus on the positive aspects of the experience here. I can of course only speak for myself, but this has been working for me.

 

[ElectricBlue]

Where are people getting 2.5 from? The midpoint between 1.0 (minimum possible score) and 5.0 (maximum possible) is 3, not 2.5.

Which sorta kinda demonstrates the problem some people have understanding any kind of scoring system, when they stumble on basic maths...

 

[onehitwanda]

Well hello.

Is that a Poisson distribution in your pocket or did we just forget to normalise things?

 

[AwkwardlySet]

There is no perfect rating system. No matter what rating system one chooses, someone will find a way to abuse it, and it won't take long before someone finds something that is unfair or unrealistic about it.
That being said, the rating system Literotica uses is just plain bad. It's too easily abused and manipulated, and it sets some silly thresholds, while the top lists make little to no sense in the way they rank stories.
There were several systems we discussed here that were clearly much better than this one but yeah, it's all empty talk. Things don't change around here.

 

[Bramblethorn]

Is that a Poisson distribution in your pocket or did we just forget to normalise things?

I do like them long-tailed.

 

[StillStunned]

I do like them long-tailed.

Are we still talking about fish?

 

[Bamagan]

Statistics for Seduction: Unlock your Cas-ANOVA

 

[Britva415]

A 2.5 rating should be the Centerpoint

Good luck getting voters to adhere to that.

 

[Bramblethorn]

Are we still talking about fish?

You somehow left off the "er's F-distribution" part of the question, but yes, absolutely.

BTW, if anybody needs a nerdy title for an I/T story, Fisher also gave us the Sexy Son Hypothesis.

 

[onehitwanda]

This thread is not ridiculous enough.

 

[StillStunned]

Oglaf makes everything better.

 

[Bramblethorn]

This thread is not ridiculous enough.

I think that may be the only Oglaf where the dwarves find somebody who's as enthusiastic about their inventions as they are.