4 ms·
Ask PG: Has the algorithm been modified?
It seems that much less votes are needed to get a high position on the front page. I see stories with 5 points reaching nearly the same position as one with 600+ points. Which would be fine actually, if the 600pts are accumulated over two days and the 5 points within 10 minutes, but the 5 points story is an hour old already.
So what I'm asking/suggesting really is:
1. Has the algorithm been tweaked or changed so that a story makes it to the homepage more quickly, or does it just seem like that?
2. Can it be changed so that a story needs more upvotes before appearing on the homepage? I think this improves the quality of homepage stories.
- iProject 14y agoCan't speak from inside knowledge, but it has been observed when this was asked in past that during slow period (middle of night, U.S. time; Sunday mornings) it takes surprisingly little activity to rise quickly to top (if only briefly).
- 46Bit 14y agoThis makes sense, really. So few votes are cast (on page #1 of the newest, often seemingly 10-per-50-posts) that people voting means something is significant.
- jQueryIsAwesome 14y agoMay be, but it lets luck play a bigger role and looks like an abusable feature.
- jacques_chester 14y agoIt's a feature that means that those of us living outside the continental USA are not disadvantaged.
- jacquesm 14y agoRight now from what I can see it takes 3 votes to get to the homepage (I checked using the Swequity link, it was at 2 and not on the homepage, after one more upvote it was), but those 3 upvotes have to be in a fairly limited timeframe to be effective. 3 points in 5 hours doesn't cut it, but 3 votes in a half hour or so is enough. How long it will stay there depends on how many votes and flags it gets after that. A flag counts as multiple downvotes.
- Gring 14y agoMaking it easier for a post to hit the frontpage might be desirable to reward people for interacting with the "new submissions" page - if it only needs 3 votes, there is a much higher likelihood of instant gratification. Might be something to build into the next algorithm.
- kami8845 14y agoIf you want to confirm/disprove your hypothesis you can use a tool I built for just this usecase: http://rewindhn.com http://rewindhn.com It also has an API if you want to play with the raw data yourself.
- danso 14y agoWow, this is excellent, the best HN-related app I've used in awhile. I obviously missed it when it was on the front page.
- jacques_chester 14y agoDuring quiet periods it's easier to hit the front page. A lot of stories get on the front page during the European daytime and get a second round of voting when the US east coast gets to work.
- irahul 14y agoPrevious discussions: http://news.ycombinator.com/item?id=1781013 http://news.ycombinator.com/item?id=1781013 http://amix.dk/blog/post/19574 http://amix.dk/blog/post/19574 The simplified algorithm is Score = (P-1) / (T+2)^G where, P = points of an item (and -1 is to negate submitters vote) T = time since submission (in hours) G = Gravity, defaults to 1.8 in news.arc The updated algorithm posted by pg has other factors, but those are related to quality control. If the same algorithm is still deployed, stories decline with time. So, during quiet periods, the new stories with far lesser number of votes rank higher than old stories with high number of votes.
- sambeau 14y agoI suspect that what you are witnessing is nothing more than the Thanksgiving holiday effect.