Though most people will check under new stories and latest stories the other story search modes seem useless such as Category and Rating. Whenever you search story by category it brings up a list of stories but there seems to be no order to the stories. And none of the stories that appear on the front pages have been updated in years. So unless one wants to search through countless pages this seems pretty silly to me. I know the old site had it in alphabitacal order but not saying this was any better either. Also stories by ratings seems pretty unfair alot of the stories on the first few pages have just one vote of one perfect five star meaning most of these stories or fresh while another story can have 99 five star votes and 1 four star vote and it will get knocked back who knows how many pages. Maybe have a minium of votes to get on the list if possable. The number one story is black ball with 1 vote, created in 2001 and no updates since it went on the new site. Stories by threads seems okay though they don't tend to be very interactive after a certain point it is just a straight shot of threads with no choice. Though I do like story by views because it means the story has to have some length to it, but not always as well interactivity to get people to come back a second time and not just jump to the new thread you enetered.
We could simply use maths to weight stories by more votes being heavier. We'll just have to cross-reference two variables: The story's average rating, and its number of votes. Then the question will be which weighs heavier: A five-star story with one vote, or a four-star story with a hundred votes. Any mathemagicians in the house? On the other hand, this would encourage people to vote if they really want to read stories based on how often people vote on it, rather than content. Omitting stories based on a lack of attention would censor stories from the results based on something as arbitrary as popularity. Step one is to verify what it is we're calculating here, so that we can represent it practically. I think this link might be the key, but I've yet to fully comprehend it myself: http://www.buzzle.com/articles/weighted-average-calculation.html I'm worried this weighting is a comparative between two specific examples, when what we need is an array that can be automatically sorted by machine. Using Rolz.org, I gathered the following mathematical example. Imagine the following are arrays of story votes. 5 1 2 3 3 5 1 2 2 2 3 3 4 4 5 5 Story A has a 5 out of 1 vote. Story B has a 14 out of 5 votes. Story C has a 31 out of 10 votes. If we just take a clear average, Story A has a 5, Story B has a 2.8, and Story C has a 3.1. Or we could just rate them by the aggregate total of each vote's score added together. That way a story with one vote of 5 could only ever remain with a score of 5, while stories of 14 or 31 exceed it. Even a story with six votes of 1 will exceed it, however, which trumps that solution. Of course, numbers that can hypothetically approach infinity will invariably lose their scope of comparison to one another, which is why we use the averages in the first place. Ultimately what we might want to achieve is to combine the average and the quantity of values calculated into that average into one sensible figure. I guess we could secretly weight them based on their total score, but keep the average as the display value. This will lead to confusion however, as people may overlook the number of votes and wonder how a story with 1 star voted on 100 times beats a story with 5 stars voted on 10 times. The issue returns to what we're calculating: We're calculating the perceived quality of a story, and the quantity of times individual readers have been compelled to vote on it. It's easy enough to abstract this concept, but simple numbers need something more concrete.
Or maybe just atleast a vote limit minium before it can be added to the highest rated list. On the first page only one story has over ten votes, same for the second page as well, and so on and so on.
I would use a Bayes estimator to rate the stories (http://en.wikipedia.org/wiki/Bayes_estimator). This can be done by a formula like Weighted Rating = (v / (v+m)) * R + (m / (v+m)) * C. R equals the average rating of the story, v the number of votes, m the minimum votes required to be listed, and C the mean vote across all stories. The weighted average just gives you the average value of all stories, or the average rating per story which is the same as the "clear average".
If I'd stayed up a bit longer, I might've found that, too. Or so I like to think, heheh. That's a lot simpler, though it seems like it'll have to do a massive amount of calculations. I'm kind of curious, actually, how many stories are on the site. And each story averages its score from each of its threads, yes? (Incidentally, I wonder how those threads created for #LINKTO apply here, but I admit I'm still not entirely cogent on that feature yet.) This can be sorted by ascending or descending, yes? I'm thinking about readers looking for stories specifically to vote on. This equation can't include stories that haven't been voted on, or it would equal out to: 1 * R + 0 * C = R So that sorting method is still dependent upon popularity. I suppose with a minimum vote threshold of 1, it would still work using long decimals, and you could simply list all non-voted stories by submission date, or something. I guess not many people would be interested in voting on a story that has no votes on it already, although somebody's got to break the mold. Um. Did I just overlook suggesting that search results should be sorted by a hierarchy of variables? So for instance, a search for score from high to low would also sort those results by votes. First, it should sort results by their rating, but then it should sort results by their votes. So all the 5-star ten-vote stories will appear by default above the 5-star one-vote stories. Of course, as we've elaborated above and Friedman has pointed out with the weighted rating formula, this still causes the core discrepency of 4-star ten-vote stories being sorted beneath 5-star one-vote stories. I suppose that I'm really just interested in knowing if the search sorties have this sort of multi-layered context sorting built in to begin with.
