2013 analysis finds Hacker News ranking relied on hidden penalties and periodic reranking
A study of the site’s front-page stories found its published formula broadly matched observed rankings, but suggested additional penalties affected about one in five articles.
A 2013 analysis of Hacker News found that the site’s published ranking formula broadly reflected observed behaviour, while additional and partly opaque penalties shaped the front page.
The analysis examined the top 60 stories over several days and estimated that about 20 per cent of front-page articles were penalised in some way. The reported penalties included measures affecting stories described as controversial, gags, lightweight posts and allegedly over-popular articles.
A severe controversy penalty was observed when a story had more comments than upvotes and reached about 40 comments. In the cases examined, articles could fall sharply in the rankings or disappear from the top 60. The analysis suspected the penalty was stronger than the exponent stated in Hacker News’ published code, but did not establish the exact formula.
The site’s underlying formula includes time decay, known as gravity, which causes scores to decline over time. However, the analysis found that stories were not continuously ranked from scratch. A story was generally reranked after receiving a vote, while one of the top 50 was reportedly selected for reranking every 30 seconds to prevent inactive stories remaining in place. Cached pages could remain unchanged for up to 90 seconds.
The findings were based on empirical observations rather than access to Hacker News’ live production system. They did not establish whether penalties were imposed by administrators, user flags, automated systems or a combination of factors. Vote counts may also have differed from actual totals because some votes were reportedly suppressed. The analysis noted that Hacker News later removed penalties for stories containing “NSA” in their titles.
