CARMELO (Career-Arc Regression Model Estimating Player Outcomes) was a player projection system developed by FiveThirtyEight. It matched current players to historical comparables based on age, size, and statistical profile to project future performance and career arcs.
CARMELO identified the N most similar historical player seasons using a nearest-neighbor algorithm across a battery of performance metrics, then weighted those comparables by similarity to project the target player's future RAPTOR scores.
One of the first public systems to provide probabilistic career projections rather than single-point estimates. Integrated aging curve research with player tracking data.
FiveThirtyEight ceased operations in 2023, ending CARMELO updates. The model was explicitly designed for projection (future performance), not retrospective evaluation of what a player did in a completed season. APEX evaluates completed seasons, not future trajectories.
CARMELO is not used in APEX for two reasons. First, it is defunct — no longer updated since FiveThirtyEight's closure. Second, it was a projection tool, not a retrospective evaluation metric, which is a different analytical problem from what APEX solves. The projection use case is a potential future APEX feature (see V2 Roadmap), not a current model component.