Crazy Uncle George's Rankings
Back to Rankings
So, you want to know what these numbers mean. Well, I'll have to admit one thing. I didn't put this system together. Duffman did since he's a statistician, and let's face it, I don't even know who's in my own cabinet. No, I don't mean the executive branch. I mean there's some weird hobo in my pantry. Anyway, I'm going to explain the system as best I can.
Now we start with one thing: server data. The server tells us which locations people accessed and how many times. Now I can't tie this to an actual file, just a name of a file. So let's say I got hot pictures of some woman named Alotta Fagina. According to my naming convention, her first picture would be named afagina01.jpg, unless there was another A. Fagina. I try to rank pictures, though, so I might have put up a picture of Alotta Fagina in a white bikini as afagina01.jpg for the month of February and changed that picture to the name afagina22.jpg for the month of March (and let's say there was no picture by this name before), while a picture of her in a black bikini became the new afagina01.jpg. So at the end of March, I'd have a download number for afagina01.jpg that would be the downloads of the white bikini picture during February plus the downloads of the black bikini picture during March, and a download number for afagina22.jpg that would be the downloads for the white bikini picture during March.
Now the question becomes how to compare different girls based on these numbers. I should also add that none of these numbers show up until the location has been downloaded 20 times. Anyway, it's not fair to compare the total downloads for girls, as some have many more pictures. At the same time, it's not fair to compare their most downloaded picture, as more people will download the same picture if a girl has fewer pictures, i.e. if they have fewer choices. So I came up with a formula, that takes the number of downloads for the most downloaded picture of a girl (again, let me note that the downloads include other pictures previously at that URL) plus 2/3 the second most for a girl plus 4/9 the third most and so on until you run out of pictures. (The old domain only rankings use 1/2, 1/4, etc.) But not every girl has been on the page for an equal amount of time, so we have to adjust for that. We have time factors for every day on the server. They basically take the total number of hits (i.e. HTML pages downloaded) the server got for its entire span and divide by the total hits the server got for the period when the girl was on the page.
Traffic for All Time by Week, after Subtracting Pages within Pages
So now we have some giant number for every girl. How does that translate to the grade you see? Well, have you ever heard of "curving" a test? Usually I was used to it meaning a flat percentage being added to every score, e.g. the highest score on a test is 80% so my 12% becomes 32%. Well, that's not exactly what curving means, as its roots are from fitting scores to a bell curve. This means that scores would be ranked and fit to a predefined spread, so someone would set the center of the curve at 75, the bottom at 50, and the top at 100 (the center and the spread can be changed, of course). So let's say you have the lowest grade in the class and it's a 72. You'd actually get bumped down to a 50. The shape of the bell means the highest number of scores are where the shape of the bell is highest. It's a way of clustering scores near the average. Anyway, we don't do that. This is sort of an in-between, the type of approach you'd use if you're, say, scoring the SATs.
Basically, these scores are standardized. The important thing here is a number called the standard deviation. You add together the squares of all the differences between the scores and the averages, divide this sum by the number of scores, and take the square root, and this is the most common measure of spread out there. So we figure out how many standard deviations above or below the average each score is, and then define the average as 75. We then define the second highest score (for the old domain only rankings, the highest instead of the second) as 100, which sets the value of how much each standard deviation is worth in our standardized scores. There, so now you know the elaborate system for generating a simple set of hotness grade numbers.