Skip to content

Latest commit

 

History

History
58 lines (44 loc) · 2.44 KB

desiderata.md

File metadata and controls

58 lines (44 loc) · 2.44 KB

Desiderata for a Trending Algorithm

An ideal trending algorithm would have the following properties. All of the ones that exist meet at least a few of these criteria, but none satisfies all of them. It could be impossible to satisfy all of them.

Which quantities should be used (more)

  • It should mostly use on-chain quantities, to reduce/remove reliance on authorities such as LBRY, Inc.

LBC

  • An increase in the total amount on a claim (bid + support_amount) should improve its trending performance
  • A decrease in the total amount should hurt a claim's trending performance
  • It should be possible for a user with a lot of LBC to use it to make a claim trend artificially.
  • On the other hand, it should not be possible for a single user or a small group of users to dominate trending for a large proportion of the time (Tom's time-locked supports and/or my delayed_ar version mitigate this).
  • A claim that receives a small amount of LBC spread over many supports should have a chance against claims with a lot of LBC.
  • However, it should be difficult or impossible, to use the previous point and a script to dominate trending for a large proportion of the time (i.e., perhaps it is possible to do this but then you have to wait before you can do it again).

Time

  • Recent changes should have a greater effect than changes that happened further in the past
  • Ideally, trending scores/ranks should be updated quite frequently, so the experience is responsive

Reposts

  • Other on-chain quantities such as reposts could, and probably should, be used, though there
  • If so, it should not be possible to artificially make a claim trend by reposting it from worthless dummy channels created for this purpose.

Trending list

  • The top trending spots should include a mix of things that are trending in an absolute sense (i.e., more popular than everything else recently) and things that are trending in a relative sense (i.e., more popular than it used to be).

  • However, it should not include things that are trending in a relative sense but performing poorly in an absolute sense (e.g., something that had 0 views, then suddenly jumped to 5, should not be high in trending).

  • In my opinion, the top 20 or so trending spots should be about a 75/25 mixture of organically popular content and promoted content, respectively.