A fixed assignment matrix maps true two-state probabilities into observed readout probabilities and integer expected counts. Exact arithmetic here means exact results for the stated model inputs; measured inputs still carry uncertainty and significant-figure limits.
highlighted = computed this step
Prediction reuses the calibrated assignment matrix
The prediction scan uses the checked matrix with first row 9/10, 1/10 and second row 1/5, 4/5.
A=[1095110154]
Three true states become three observed readout budgets
The scan pushes a true-zero state, an even mixture, and a true-one state through the same assignment matrix with the same shot total.
The true-zero row begins with probability 1 on zero, but the observed one-read probability is 1/10.
109+101=1
Prepared one uses the second calibration row
The true-one row begins with probability 1 on one, and produces 16 expected one-read counts out of 20 shots.
51+54=1
Prediction closure is finite-count arithmetic
This closure does not claim detector drift, deconvolution, or hardware fidelity. It only shows how a calibrated assignment matrix changes finite expected counts.