Is allocation.FACTORS banded on the board on the quantity it was fitted on, and if not, what does the mismatch cost?

Owner lane: calibration

Question

Is allocation.FACTORS banded on the board on the quantity it was fitted on, and if not, what does the mismatch cost?

Prediction

Band disagreement 5-15% of RB/WR/TE rows (H-212's own), low end in R2 and higher in R1. Walk-forward MAE contrast SERVED-BASE nil pooled (|d| < 0.01 pts/wk); on changed-band rows SERVED worse by a positive amount whose interval may span zero. REFUTED if disagreement < 1% of rows OR the multiplier change nets to a rounding error (pooled |d| < 0.002 AND the changed-row interval spans zero).

Finding and verdict

TESTED. The board's FACTORS band input is not the unit FACTORS was fitted on: live today 16.2% of carry bands and 25.9% of target bands differ from the target-row fitted unit (moving projections 0.234 [0.192, 0.278] pts/row, RB 0.40), walk-forward 3.5-5.8% in season and 12-24% in weeks 1-3. Fed the served unit walk-forward, fantasy MAE moves +0.0008 [-0.0038, +0.0054] pooled and +0.017 [-0.027, +0.060] on the 1,206 changed rows: MEASURED NOTHING on accuracy, with a shuffle control that clears zero only narrowly.

Reasoning

A correction fitted on one input and served another is the D-71 defect class; measured, the mismatch is a movement the chain absorbs rather than an error it adds, on this population and metric.