Hacker News的文章排名基于一套复杂的算法体系，核心公式综合考量投票数、提交时间和多种惩罚因子1。该算法采用重力系数1.8作为时间衰减指数，使得新提交的内容获得排名优势，但随时间推移权重逐渐下降1。
该平台对特定内容类别实施了自动惩罚机制1。含有"NSA"字样的标题会被自动降权至0.4系数，相当于投票权重下降70%1。同时，多个热门新闻网站域名如arstechnica.com、businessinsider.com和github.com等被设置了0.25至0.8不等的惩罚系数1。此外，当文章评论数超过40条且超过投票数时，会遭受(votes/comments)³的争议性惩罚因子1。
根据2013年11月11日的数据分析，这些隐藏的惩罚机制影响范围广泛：首页文章中约20%遭受过惩罚，第二页文章的受罚比例更高达38%1。为维持排名动态性，平台平均每30秒对前50个故事中的一个进行随机重新排序，同时页面缓存周期为90秒1。研究者通过逆向工程Arc源代码发现，实际的惩罚系统远比公开代码所显示的内容更为复杂1。
An analysis of Hacker News' article ranking algorithm reveals a complex system that extends far beyond simple vote counting. The ranking formula incorporates the number of votes received, the time an article was submitted, and a gravity coefficient of 1.8 that causes older stories to decay in prominence over time 1. Beyond these basic components, the platform employs multiple hidden penalty mechanisms that significantly suppress the visibility of certain articles.
The system applies several categories of automatic penalties that reduce articles' effective ranking power. Articles generating substantial discussion face a "controversy penalty" when they accumulate more than 40 comments and receive more comments than votes, which applies a penalty factor of (votes/comments)^3 1. Domain-based penalties affect popular news sources, with sites including Ars Technica, Business Insider, and GitHub receiving penalty coefficients ranging from 0.25 to 0.8 1. Title-based penalties also exist—articles containing "NSA" automatically receive a 0.4 penalty coefficient 1. The impact of these penalties is substantial: a coefficient of 0.4 reduces voting weight by approximately 70 percent, while a coefficient of 0.1 causes stories to decay 3.6 times faster 1.
An examination of Hacker News data from November 11, 2013 demonstrated the widespread application of these penalties across the platform 1. Approximately 20 percent of articles appearing on the homepage had been penalized, while 38 percent of articles on the second page suffered some form of penalty 1. The ranking system continuously updates, with one story from the top 50 refreshed on average every 30 seconds, and pages cached for 90 seconds 1. Investigation into the platform's Arc source code uncovered that the actual penalty system operating in practice is considerably more intricate than what appears in publicly available code 1.
评论
还没有评论，欢迎留下第一条。