Description
Compare predicted and reference single-class squared probability errors.
Prediction metrics depend on the observation definition, reference forecast, denominator sign and aggregation. A single error contribution differs from a mean score or normalized skill measure. Interval midpoints, squared probability errors and seasonal-denominator labels must retain the exact convention before models are compared.
Compare predicted and reference single-class squared probability errors.
When to use this tool
Use Brier Reference Class Contribution when you already have classProbability, observedClass, referenceProbability and need to compare predicted and reference single-class squared probability errors.
How to use the tool and how it works
- Enter classProbability, observedClass, referenceProbability. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read classSquaredError, referenceSquaredError, improvement. Use the output convention below.
Require both probabilities in [0,1]; observation is 1 when observedClass else 0. Return each (probability-observation)² and reference error minus class error as improvement.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| classProbability | Class Probability | Required. Finite number; minimum 0; maximum 1. |
| observedClass | Observed Class | Required. Boolean. |
| referenceProbability | Reference Probability | Required. Finite number; minimum 0; maximum 1. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| classSquaredError | One (classProbability-observedClass)^2 contribution; not an aggregated multiclass Brier score. | Required. Finite number. |
| referenceSquaredError | Reference Squared Error | Required. Finite number. |
| improvement | Reference contribution-current contribution; positive favors the current probability for this one outcome. | Required. Finite number. |
{
"input": {
"classProbability": 0.5,
"observedClass": true,
"referenceProbability": 0.25
},
"output": {
"classSquaredError": 0.25,
"referenceSquaredError": 0.5625,
"improvement": 0.3125
}
}Limitations and assumptions
- Positive improvement means smaller error than the supplied reference for this observation; this is neither an aggregate multicategory score nor a normalized skill ratio.
- Inputs must satisfy both the field types and the cross-field conditions above. Invalid input causes an error; a successful false flag or documented placeholder is a different result.
- Numeric calculations use binary floating point unless exact decimal-integer arithmetic is explicitly stated. No hidden measurement tolerance or decimal rounding is added; nonfinite or out-of-contract results are rejected. Numeric signed-zero results are exported as zero.
Alternative or Complementary methods
Calculate the complete documented metric over the intended observations and reference baseline. Review missing/zero actuals, aggregation weights and outcome coding; one contribution or source label does not establish overall forecasting quality.
References
Similar or alternative tools
- Mape Absolute Vs Signed Actual Policy
Calculate one absolute prediction error under a chosen actual-value denominator convention.
- Scaled Error Seasonal Denominator Tag
Qualify an explicit scaled-error denominator by seasonal-period wording and source.