3 ms·
The way I do it is points divided by age in hours (plus two) to the 1.25th power. This is probably not optimal, but it's simple. (= gravity* 1.25 timebase* 12
by pg 19y ago
The way I do it is points divided by age in hours
(plus two) to the 1.25th power. This is probably not optimal, but it's simple.
(= gravity* 1.25 timebase* 120)
(def hotness (s (o gravity gravity*))
(/ (- (realscore s) 1)
(expt (/ (+ (story-age s) timebase*) 60) gravity)))
I avoid reranking all the stories by keeping them in
sorted order, and whenever I generate a frontpage for
someone logged in, reinserting one randomly
chosen story near the top (where all the change is).
I also reinsert any story that gets voted on.
- ivankirigin 19y agoInteresting. Right now, I have no data, making testing difficult -- especially for anything that will change hotness depending on distribution of mods in a given time. It's been something I've been thinking about for a bit, and want to get it out of my head to focus on building something that people use so I can actually get the data.
- darragjm 19y agoWhy not just start with a simple hotness algorithm and evolve it after data starts coming in?
- ivankirigin 19y agoThat's exactly what I'll do.
- run4yourlives 19y agoThanks for sharing that Paul. Most people guard that knowledge like it's the lost grail. Kudo to you for being open about it.
- ivankirigin 19y agoIndeed