YOUR RANKING
How ranking works
A structured way to put your Taylor Swift favorites in your own order.
Each ordered set becomes evidence
Start with songs you know and place each small group in your own order. Every completed set records which songs you preferred within that set. Maple Lattes combines those ordered sets into an estimated ranking with a regularized Plackett–Luce model. Regularization limits extreme estimates when a song has little comparison evidence, so a small number of answers does not produce an outsized score.
Later sets focus on unresolved places
The first pass introduces selected songs. After that, the comparison schedule concentrates on places the current order has not settled, first around the Top 10 and then the Top 50. Top 10 sets are smaller, with up to five songs; other sets can include up to eight. When answers conflict, the latest complete answer takes priority when forming your displayed order.
You keep direct control
You can set unfamiliar songs aside, review the proposed Top 10, or use Move this song to compare one favorite with nearby songs at a new place. You can also revisit a group, save dated editions, choose Original or Taylor’s Version recordings for matched songs, and use rediscovery before bringing a song back into your ranking. These actions add your current choices without asking you to rebuild the rest of the list.
Method background and limits
Plackett–Luce models and adaptive subset ranking have an established statistical foundation. For background, see Turner et al., Modelling rankings in R: the PlackettLuce package, and Saha and Gopalan, Active Ranking with Subset-wise Preferences. Maple Lattes combines these ideas with its own comparison schedule and review tools. The statistical foundation is established; this particular workflow has not yet been validated in a real-user study. It reflects the current product choices and may change.