Appreciate the collab @TipOff and the beautiful graphics want to add a couple more words on LOKI: LOKI is a predictive metric built on the philosophy of possession-by-possession fundamentalism. Tested out of sample, it matches or beats EPM, DARKO, and LEBRON in predicting games and team win percentages using preseason ratings (2020–21 to 2025–26). Quantifying the answer to the following question: "Given the limited number of possessions, in an environment where every action within a possession can be quantified, whether tangibly (EPV) or abstractly (RAPM), what is the incremental value a player generates for his team?" One of the key tenets of LOKI was to stay decomposable rather than being an opaque number. Building on the great RAPM decomposition work of @databallr and 4 factor framework of @DeanO_Lytics, can we take the opaque total number and contextualize it into the channels through which a player actually affects the game? If you're curious to learn more, my DMs are completely open. Here's some reading, and here's a mini paper I wrote on the development of ARC v3, the descriptive metric it was built on: xitter.cf/Starro____/status/2108…