Weight of stars?

[Scandia]

Somebody just went around giving 1-star ratings to nearly all my videos. It decreased the rating of my best video from 3.56 (which I was EXTREMELY proud of) to 2.96 (not even a 3.0). Plus 3 others that had around 3.0 are now quite below that.

NOBODY specific is being accused of this.

But why would one lousy star giver reduce the rating of one high ranking video so much? One that has such good opinions and that has always been over 3.0 (in fact, over 3.5) since the very beginning?

Needless to say, I am VERY upset about this.

[AbsoluteDestiny]

So someone doesnt like your videos... it's not the end of the world. If you're going to get annoyed about something as petty as this then perhaps you should only make videos for private enjoyment.

[Scandia]

You see, I am ferociously competitive. I was the kid who cried because of anything lower than an A in school. Once I cried because of a 96. I was even nominated in my fraternity as "most likely to put a Smith and Wesson in her mouth due to a B+". I also love positive attention and recognition.

I was just a bit confused at the sudden drop in rating.

I am sorry if this has caused any problems. I hope I am not disliked due to my inquiry. I can be a pest, but it is NOT my intent to annoy people on purpose.

[trythil]

But why would one lousy star giver reduce the rating of one high ranking video so much?

1/N, where N is the number of samples in the arithmetic mean, i.e. stars.

[Scandia]

1/27 is much less than .5. And the video had about 27 stars before it got that lousy rating.

Anyhow, forget it. I do not want to be consider a spoiled brat or whiner in the AMV community.

[trythil]

1/27 is much less than .5. And the video had about 27 stars before it got that lousy rating.

I don't think you caught the point. Perhaps the derivative will be a better way to describe this.

d/dn (1/n), of course, is -(1/n^2). This function asymptotically approaches -infinity as it approaches zero, and asymptotically approaches zero in the limit of n to infinity. Therefore, small values of n exhibit large changes, and large values of n exhibit small changes.

The arithmetic mean is (1/n)*sum(xi, i=1..n). (1/n) is directly proportional to its derivative: for small n, 1/n is large; for large n, 1/n is small. Therefore a small number of samples (i.e. ~27) means that more "weight" is distributed across each sample than in a large number of samples.

If n is the number of stars you have, then it is clear from the behavior that at around x=27, a very low summand can still have quite an effect. The examples ((27*5)+5)/28 (=5.00) and ((27*5)+1)/28 (~= 4.85) should demonstrate this concretely.

In other words, you're railing against mathematics. I don't know -- if you're really angry over this incredibly small issue, go travel back in time and kill Leibniz and Newton, or something.

[Scandia]

I haven't used calculus in a while. Dornkirk is cool. I am not as upset anymore. There are many more important things in life.