Roger predicted the quarter before the accountant filed. The prediction was frozen.
This is the measurement of the gap — the control the whole Tax Mode build was designed around.
RECONCILIATION RUN
Every stage below actually executed to produce this page — nothing here is a mock.
ROGER · FROZEN
€0.00
VS
ERIK NEVE · FILED
€0.00
HEADLINE DELTA
€0.00
WHY THE HEADLINE LIES
Each bar is a place where Roger's method and Erik's method genuinely disagree.
They very nearly cancel. That near-cancellation is the whole reason a headline match proves nothing.
PRE-REGISTERED HYPOTHESES
Written down on 23 Jul, before Erik filed — so the scoring can't be retrofitted.
DISCOVERED DIVERGENCES
Not predicted. Surfaced by the diff itself — honestly labelled as such.
PER-CATEGORY LEDGER
Claimable VAT per category. Roger's reverse-charged lines net to €0 against the refund,
which is how they're compared here.
Category
Erik
Roger
Delta
ENCODE INTO recon.py
The point of the exercise: turn the accountant's method into Roger's rules before Q3.