WHY MATHEMATICS?

Why Mathematics? | Energy Score for Multivariate Forecasts: Euclidean Distances, Ensemble Spread, Joint Outcomes and Dependence Limits

How one geometric score extends CRPS to vectors, with a complete two-dimensional calculation, scaling choices and honest dependence checks.

Why mathematics is important becomes easier to see when a real decision depends on it. This guide follows the mechanism from definitions and dimensions to a worked calculation, validation, practical learning and responsible use. It keeps factual evidence separate from interpretation and treats every illustrative scenario as an example rather than a promise of results.

Choose a reading route

Represent each forecast and outcome as a vector
Measure member-to-observation Euclidean distance
Subtract half the member-to-member distance expectation
Add marginal and dependence diagnostics where the total is silent

Chapter map

1. Vector target
2. Member one
3. Member two
4. First term
5. Pair self-distance
6. Cross-distance
7. Pair mean
8. Spread credit
9. Energy score
10. Scaling
11. Dimension
12. Dependence
13. Marginals
14. Fair version
15. Uncertainty
16. Claim

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

1. Vector target

The useful question is not only how to calculate, but what the calculation assumes. Define two coordinates.

A concrete situation

Use y=(2,3).

What the mathematics does not license

Coordinate order is part of the target. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Label dimensions. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

2. Member one

A small example makes the mechanism visible. Use x1=(1,2).

A concrete situation

Distance to y is square root of 2.

What the mathematics does not license

Units affect Euclidean length. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

State scale. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

3. Member two

This is where modelling choices begin to affect evidence. Use x2=(3,4).

A concrete situation

Distance is square root of 2.

What the mathematics does not license

Symmetry here is illustrative. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Keep case label. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

4. First term

The central mathematical idea is concrete. Average two observation distances.

A concrete situation

Result is square root of 2.

What the mathematics does not license

Member probabilities must be explicit. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show mean. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

Did You Know? In this chapter, the most important number is useful only together with its construction rule. Changing scaling, sample membership or validation design can change the numerical answer without changing the arithmetic.

5. Pair self-distance

Here the geometry becomes operational. Compute x1 to x1 and x2 to x2.

A concrete situation

Both are zero.

What the mathematics does not license

Self-pairs appear in empirical form. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Keep matrix. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

6. Cross-distance

This step deserves a careful pause. Compute x1 to x2.

A concrete situation

Distance is square root of 8.

What the mathematics does not license

Use the same norm and scaling. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Verify squares. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

7. Pair mean

The useful question is not only how to calculate, but what the calculation assumes. Average four ordered distances.

A concrete situation

Result is square root of 2.

What the mathematics does not license

Unordered-pair formulas need rescaling. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

State convention. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

8. Spread credit

A small example makes the mechanism visible. Take half the pair mean.

A concrete situation

Result is square root of 2 over 2.

What the mathematics does not license

Spread is credited, not rewarded without limit. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Explain balance. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

9. Energy score

This is where modelling choices begin to affect evidence. Subtract spread credit.

A concrete situation

ES≈.7071.

What the mathematics does not license

Lower is better under loss orientation. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Label direction. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

10. Scaling

The central mathematical idea is concrete. Standardise coordinates only with a fixed rationale.

A concrete situation

Record reference scales.

What the mathematics does not license

Data-driven post-outcome scaling changes target. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Freeze rule. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

11. Dimension

Here the geometry becomes operational. Keep vector length fixed across models.

A concrete situation

Score identical target spaces.

What the mathematics does not license

Extra coordinates change distance geometry. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Match dimensions. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

12. Dependence

This step deserves a careful pause. Inspect correlation and joint ranks separately.

A concrete situation

Use a variogram companion.

What the mathematics does not license

ES may have limited discrimination for dependence errors. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Add evidence. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

13. Marginals

The useful question is not only how to calculate, but what the calculation assumes. Calculate univariate scores per coordinate.

A concrete situation

Find local weaknesses.

What the mathematics does not license

Good joint total can hide one poor margin. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show components. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

14. Fair version

A small example makes the mechanism visible. Adjust pair denominator for sampled ensembles when justified.

A concrete situation

Use M(M-1).

What the mathematics does not license

Fairness assumption must be declared. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Compare forms. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

15. Uncertainty

This is where modelling choices begin to affect evidence. Use paired, dependence-aware score comparisons.

A concrete situation

Report sampling variation.

What the mathematics does not license

One vector is not evidence of superiority. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Aggregate. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

16. Claim

The central mathematical idea is concrete. State vector, norm, scaling and ensemble form.

A concrete situation

Teaching ES is about .7071.

What the mathematics does not license

Do not claim dependence is proven correct. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Bound conclusion. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

A seven-day practice lab

Day 1: define the decision, observation unit, X variables and Y or reference condition.
Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
Day 3: build a transparent baseline and write what it assumes.
Day 4: fit the main method inside a training-only pipeline; record every parameter.
Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
Day 6: compare validation metrics with the practical workload created by errors.
Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

Which part of the model is a mathematical definition, and which part is a choice?
What small calculation would catch a sign, scale or dimension error?
What evidence comes from training, validation and the untouched test?
What benign condition could look like a fault or prediction failure?
How would the answer change for another response, asset, school project or time period?
What should a human review before acting on the score?

A method-specific audit before deployment

The energy score generalises the one-dimensional CRPS to a vector outcome. It compares the average Euclidean distance between forecast draws and the observed vector with half the average distance between two independent forecast draws. The first term rewards closeness to the realised joint outcome; the second credits appropriate ensemble spread. Gneiting and Raftery develop the score within the theory of proper scoring rules and negative-definite kernels.

Let an ensemble contain x1=(1,2) and x2=(3,4), with observed vector y=(2,3). Each member is distance square root of 2 from y, so the first term is square root of 2. The ordered member-to-member distances are 0, square root of 8, square root of 8 and 0. Their mean is square root of 2, and half of that is square root of 2 divided by 2. The energy score is therefore square root of 2 divided by 2, about .7071.

A vector distance depends on units and scaling. Combining temperature in degrees with revenue in dollars without standardisation can let the larger numerical scale dominate. The energy score is proper, yet finite samples may have limited power to distinguish forecasts with different dependence structures. A low score does not prove that correlations, tail co-movement or every marginal distribution are correct.

A responsible evaluation states the target vector, coordinate order, units, scaling, norm, exponent if not one, ensemble weights, missing-vector rule and aggregation. It reports marginal scores, correlation diagnostics or a variogram score beside the energy score. Students can transfer the geometry to multi-site weather, multi-sensor engineering and joint demand forecasts while keeping the distinction between closeness, spread and dependence.

Frequently asked questions

What does the energy score evaluate?

A joint predictive distribution for a vector outcome.

Why subtract forecast spread?

The kernel form balances closeness with appropriate dispersion.

Why does scaling matter?

Euclidean distance can be dominated by coordinates with larger numerical units.

Is ES the multivariate CRPS?

It is a widely used multivariate generalisation and reduces to CRPS in the univariate beta-one case.

Does a low ES prove correlations are right?

No. Dependence misspecification can be difficult to detect.

What must be reported?

Vector definition, units, scaling, norm, ensemble form, marginals, dependence checks and uncertainty.

Primary source and fact-check note

Tilmann Gneiting and Adrian E. Raftery, “Strictly Proper Scoring Rules, Prediction, and Estimation,” Journal of the American Statistical Association 102(477), 359–378 (2007), DOI 10.1198/016214506000001437. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Extended laboratory: score-specific validation and responsible use

Target audit

Freeze the predictive object, outcome vector or quantile curve, case set, units, horizons and preferred score direction before seeing comparative results.

Formula audit

Write every expectation, distance, indicator, quantile score, exponent, weight and denominator explicitly; identify ordered versus unordered pair conventions.

Computation audit

Reproduce the teaching arithmetic independently and test the implementation on identical members, perfect forecasts, constant vectors and one changed coordinate.

Weight audit

Predeclare percentile weights, coordinate scaling, pair weights or member probabilities; explain their decision meaning and units.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Support audit

Inspect quantile crossing, sparse tails, ensemble size, missing members, missing vector coordinates and weakly sampled pair expectations.

Companion audit

Report an unweighted or marginal score and calibration evidence so regional, geometric or dependence gains cannot hide deterioration elsewhere.

Uncertainty audit

Use paired comparisons and resampling that respect serial, spatial, horizon and forecast-origin dependence.

Final claim

Describe exactly which forecast feature the score evaluates; never turn one total into proof that the whole predictive distribution is correct.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Vector target

Begin with this mechanism: Define two coordinates. Recreate the concrete case: Use y=(2,3). Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Coordinate order is part of the target. Complete the transfer step: Label dimensions. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Member one

Begin with this mechanism: Use x1=(1,2). Recreate the concrete case: Distance to y is square root of 2. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Units affect Euclidean length. Complete the transfer step: State scale. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Member two

Begin with this mechanism: Use x2=(3,4). Recreate the concrete case: Distance is square root of 2. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Symmetry here is illustrative. Complete the transfer step: Keep case label. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: First term

Begin with this mechanism: Average two observation distances. Recreate the concrete case: Result is square root of 2. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Member probabilities must be explicit. Complete the transfer step: Show mean. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 5: Pair self-distance

Begin with this mechanism: Compute x1 to x1 and x2 to x2. Recreate the concrete case: Both are zero. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Self-pairs appear in empirical form. Complete the transfer step: Keep matrix. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Cross-distance

Begin with this mechanism: Compute x1 to x2. Recreate the concrete case: Distance is square root of 8. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Use the same norm and scaling. Complete the transfer step: Verify squares. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Pair mean

Begin with this mechanism: Average four ordered distances. Recreate the concrete case: Result is square root of 2. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Unordered-pair formulas need rescaling. Complete the transfer step: State convention. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Spread credit

Begin with this mechanism: Take half the pair mean. Recreate the concrete case: Result is square root of 2 over 2. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Spread is credited, not rewarded without limit. Complete the transfer step: Explain balance. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 9: Energy score

Begin with this mechanism: Subtract spread credit. Recreate the concrete case: ES≈.7071. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Lower is better under loss orientation. Complete the transfer step: Label direction. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Scaling

Begin with this mechanism: Standardise coordinates only with a fixed rationale. Recreate the concrete case: Record reference scales. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Data-driven post-outcome scaling changes target. Complete the transfer step: Freeze rule. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Dimension

Begin with this mechanism: Keep vector length fixed across models. Recreate the concrete case: Score identical target spaces. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Extra coordinates change distance geometry. Complete the transfer step: Match dimensions. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Dependence

Begin with this mechanism: Inspect correlation and joint ranks separately. Recreate the concrete case: Use a variogram companion. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: ES may have limited discrimination for dependence errors. Complete the transfer step: Add evidence. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 13: Marginals

Begin with this mechanism: Calculate univariate scores per coordinate. Recreate the concrete case: Find local weaknesses. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Good joint total can hide one poor margin. Complete the transfer step: Show components. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Fair version

Begin with this mechanism: Adjust pair denominator for sampled ensembles when justified. Recreate the concrete case: Use M(M-1). Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Fairness assumption must be declared. Complete the transfer step: Compare forms. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Uncertainty

Begin with this mechanism: Use paired, dependence-aware score comparisons. Recreate the concrete case: Report sampling variation. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One vector is not evidence of superiority. Complete the transfer step: Aggregate. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 16: Claim

Begin with this mechanism: State vector, norm, scaling and ensemble form. Recreate the concrete case: Teaching ES is about .7071. Write coordinate units, target vectors, ensemble vectors, norms, observation distances, pairwise distances, spread credit and energy score with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not claim dependence is proven correct. Complete the transfer step: Bound conclusion. Change exactly one tau level, quantile, quadrature weight, ensemble member, ensemble size, member probability, vector coordinate, scale factor, norm, p order, pair weight, threshold, forecast case or missing-data rule. Explain which mathematical object changes, which score term moves, which ranking survives and which claim must weaken. Keep target definition, calculation, aggregation, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the full arithmetic and once to a parent with the real decision and original units. Add a counterexample involving quantile crossing, post-result weighting, unequal ensemble sizes, nonexchangeable members, scale domination, common bias, identical marginals with different dependence or a good regional score with poor global performance. State the extra evidence that would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.