Scenario calibration and bracket objectives

How to read this

Expected SCN-01 score means the average ESPN-style bracket score across many saved simulated tournaments. It is not a guaranteed score and not the same thing as pool-win probability.
Direct predictions are game-level probabilities for specific matchups. Simulations are bracket-level tournament paths built from those probabilities.
This project used 12 expanding-window historical folds for model evaluation, not tournament simulations. The saved bracket layer used 10,000 FM-01 tournament simulations and 10,000 FM-02 sensitivity simulations. BP-chalk is deterministic, not simulated.
A historically credible scenario in this project means the implied upset volume, favorite survival, Final Four concentration, and champion concentration look structurally plausible against saved historical benchmarks.
What changed after enrichment: this section now reads from full saved per-game Stage 7 traces, so the plausibility check is grounded in the complete Monte Carlo record rather than the earlier lightweight sample layer.
Bottom line: the enriched traces confirm the same directional read as before, but more strongly. BP-01 is coherent and usable, yet clearly too chalky in the Round of 64 and still concentrated among a small set of elite teams.
Objective split: single-entry = the one bracket to submit; expected-score = maximize average SCN-01 points; joint-probability = prefer coherent full-bracket paths rather than isolated pick edges; portfolio = multiple coordinated entries, which remains absent in the saved output family.