I am working on a website where users can share jokes. It uses a modified version of the HN ranking algorithm that allows downvotes. Do you think the implementation is usable?

Here is how jokes are ranked:

(p-1)/(t+2)^(1.8s - (s-1)0.85)

Where:

p = votes_for - votes_against

t = time since jokes was posted, in hours

s = 1 if p-1 > 0

s = 0 if p-1 = 0

s = -1 if p-1 < 0

Essentially if s = 1, then it works the same way as the HN algorithm, and if s = 0, then the numerator and the final score are also 0.

If s = -1, then the formula changes to: (p-1)(t+2)^(0.1)

Which looks like this: http://www.wolframalpha.com/input/?i=%28-1%29%28t%2B2%29^%280.1%29+from+0+to+24

So when p-1 is negative, the score will gradually decrease with time, with the rate of decrease slowing down.

Do you think this is a good approach to downvotes? Also should I be taking the ratio of upvotes to downvotes into account, or just the difference?

I'm also thinking that maybe the rate at which jokes with negative scores loose points should increase with time, instead of decrease. That way a joke which got downvoted right after submission will have a higher chance of being seen by people and upvoted, where as a joke that's been in the negatives for a long time should get buried further down.

The website by the way is www.QuickHaha.com