Skip to content

Latest commit

 

History

History
63 lines (48 loc) · 2.04 KB

ranking.md

File metadata and controls

63 lines (48 loc) · 2.04 KB

tlks.io : raking algorithm

Items on tlks.io are scored based on its upvote score, the time since the item was submitted, and various penalties using the following formula:

score = (points) / (hours + timebase) ^ gravity

Where:

  • points is the number of an item has been upvoted
  • hours is the number of hours since the item has been published
  • timebase is the amount of hours that we want our talk not being penalized by the gravity factor, by default 0.
  • graviy is an exponent factor, by default 1.8, used as a time penalty.

So its basically votes divided by an age factor.

Because the item time has a larger exponent than the votes, an item's score will eventually drop to zero, so nothing stays on the front page too long. This factor is known as gravity.

On this release, tlks.io ranking algorithm do not implement any extra factor to boost or penalty the final score.

TODO

  • Support for homepage promotion thresolds.

    • if an item reaches an amount of score should be promoted to the homepage

  • Support for specicif situation boosts.

    • Support for point boosts.
    • Support for extra and final result boosts.

  • Support for specific situation penalties.

    • Support for point penalties.
    • Support for extra and final result penalties.

Enhanced formula:

# boosts & penalties
f = (f1 * f2 * f3 ... * fn)
ef = (ef1 * ef2 * ef3 ... * efn )
# enhanced raking calculation
score = ( (points) ^ f ) / ((hours + timebase) ^ gravity) * ef

Where :

  • f1, f2 ... fn are penalties directly applied on points.

    • Floats between 0 and 2
    • Example: Apply a 1.1 boost if an item has good metadata.
    • etc ...
    • Final penalty is calculated multiplying all falctors to get a final one.

  • ef1, ef2 ... efn are penalties applied on the final result.

    • Floats between 0 and 2
    • Example: Apply a 1.1 boost if an item has a lot of comments.
    • Example: Apply a 0.7 penalty if an item does not have a suggested field.
    • etc ...
    • Final penalty is calculated multiplying all falctors to get a final one.