Layer Stack
The goal is to keep the player valuation layer separate from the team aggregation layer so the model stays readable to basketball operations users.
Read In Order
1. DelQuant raw projection
Player output starts in the projection engine. Age, usage, durability, and similar inputs are already baked in.
2. Teammate-neutral normalization
Adjust the player feed so a player is not penalized or inflated by surrounding roster quality.
3. FMV pricing
Apply basketball-ops weights and a cap-constrained market simulation to produce salary-equivalent value.
4. FMVW bridge
Aggregate value at the roster level and calibrate that total against wins.
5. Team vs team
Apply HBB to roster aggregation for comparisons like Hawks vs Knicks.
6. Totals and proxy read
The live comparison view now shows FMV, wins-equivalent, HBB factor, HBB wins, proxy delta, proxy-adjusted wins, and the proxy readout side by side so the arithmetic stays visible.