From a list to comparisons
If a listener places Song A above Song B, that list contributes a preference for A over B. The same conversion is made for every pair of songs included in that submission.
Long lists do not overpower short lists
A complete catalog produces far more pairs than a short ranking. ListPipe divides each submission’s total pairwise weight across its comparisons, so one submitted list contributes the same overall weight regardless of length.
One active vote per person and artist
A signed-in listener can maintain one active community submission for each artist. Submitting an updated list replaces the listener’s previous submission rather than adding another vote.
Ratings and positions
The model begins around a neutral rating of 1,000. Songs rise or fall based on their weighted results against other songs. A light statistical prior stabilizes sparse early results. The visible leaderboard position is the order produced by those ratings.
Partial rankings
A song absent from a person’s list is not automatically treated as disliked. Partial rankings compare only the songs that listener included. As participation grows, broader and more complete lists create a better-connected community result.
Method version
The current leaderboard method is regularized-bradley-terry-v1. If the calculation changes materially, ListPipe will identify the new version and explain the change here.