Use the predefined rule
At the fictional June decision, growth is -0.05 points and inflation change is 0.40 points. The common boundary half-width is 0.1 around separate zero thresholds. Inflation clearly exceeds its upper boundary, but growth does not lie outside its band. The combined result is Boundary because the definition requires both signs to be resolved.
The rule was not fitted to a real historical sample. Its values are teaching choices, and the entire ledger was constructed on 16 October 2026. This is useful precisely because the arithmetic is inspectable: the example can show what an abstention rule does without pretending that 0.1 is an empirically optimal threshold for any market.
Measure distance without inventing confidence
The absolute distances to the two central thresholds are 0.05 and 0.40. The smaller distance is 0.05. Subtracting the 0.1 half-width gives a signed boundary margin of -0.05 points. A negative margin means at least one feature lies inside the abstention band; it is not a negative probability and it says nothing about expected return.
Distance has meaning only under the declared units and transforms. Multiplying the growth score by ten without changing its threshold and band would alter classifications for purely mechanical reasons. A comparison should therefore freeze the feature definition as well as its numerical threshold. Statistical confidence needs a separate model and calibration evidence.
Compare a revised pair
The second fictional vintage replaces growth with 0.20 and inflation with 0.35. The minimum central distance becomes 0.20, leaving a positive margin of 0.10 beyond the band. The latest label is G+ / I+. Those values were notionally ingested in March 2025, so they cannot make the July 2024 information-set result more decisive.
An investor-facing narrative could be tempted to call the newer quadrant confirmation of an earlier view. The record does not support that claim. The earlier rule abstained, the later input pair differs, and neither record is a forecast. Keeping the original Boundary label is an important part of describing what the method actually does under ambiguity.
Test parameter sensitivity separately
Reducing the band to zero while retaining the original scores produces G- / I+, since -0.05 is then negative and 0.40 positive. That is a definition change, not a data revision. The sensitivity workbench compares such scenarios against the fixed baseline while holding each decision's eligible information unchanged.
Both analyses are useful when reported separately. Revision sensitivity asks how labels differ under later inputs; threshold sensitivity asks how definitions differ under the same inputs. Combining them into one success percentage would obscure their causes. This case produces no allocation instruction, trading return, probability estimate or evidence about a provider's model.
Inspect the comparison
| Scenario | Growth / inflation | Half-width | Label |
|---|---|---|---|
| Original eligible pair | -0.05 / 0.40 | 0.10 | Boundary |
| Later fictional revision | 0.20 / 0.35 | 0.10 | G+ / I+ |
| Original pair, alternative rule | -0.05 / 0.40 | 0.00 | G- / I+ |
Primary references
References support the stated methods and data conventions. Original fictional examples are not observations from the cited sources.