Skip to main content

Recently Popular (Rule-Based)

Description​

The Recently Popular policy ranks items based on popularity but applies a time decay function, giving more weight to recent interactions. Older interactions contribute less to the score. This balances overall popularity with freshness. Specific decay formulas (e.g., inspired by Hacker News or Reddit) can be selected via the mode.

Policy Type: recently-popular Supports: scoring_policy

Configuration Example​

scoring_policy_trending_recently_popular.yaml
policy_configs:
  scoring_policy:
    policy_type: recently-popular
    mode: "hackernews"            # Decay algorithm ('hackernews', 'reddit', potentially others)

References​

  • Inspired by algorithms used in practice (Hacker News, Reddit). Formulas often discussed online but may not have formal publications.