PYTHON · NUMPY

The grade, not the number

Locomotion is metered and is 68% of the joules. Sensing is a stand-in. What grade is the total?

Attach a provenance grade to every figure in a composed claim, and apply the weakest-part rule rather than averaging.

01Challenge

You are handed an energy figure for a complete task: locomotion metered on a dynamometer, sensing taken from a datasheet, compute estimated from a similar model. Decide what grade the total carries.

Predict firstLocomotion is metered and is 68% of the joules. Sensing is a stand-in. What grade is the total?

02Model

A composed energy figure is only as good as its worst part, and the instinct to average is what destroys it. If locomotion is metered, sensing is a datasheet number for a component that may not be the one installed, and compute is an estimate borrowed from a different model, then the total is not two-thirds measured. It is an estimate, because an estimate anywhere in the sum contaminates the sum. The Institute's OER/1 receipt schema makes this a rule rather than a virtue: provenance axes keep modelled, stand-in and metered figures permanently distinguishable, and a composed receipt takes the weakest grade among its parts, never the average. The rule is deliberately unforgiving because the alternative rewards padding a claim with one good measurement.

03Guided practice

Give every component a grade, then compute the total two ways: by the weakest part, and by a naive average of grades treated as scores. Compare what each says about the same numbers.

04Feedback

PASS WHEN You graded a total by its weakest part and saw what averaging would have claimed instead. The largest, best-measured term did not rescue the composition, which is the point of the rule.

If your run fails, check:

composed_grade should come from the part with the lowest RANK, not from the largest J.
total_J is a plain sum; the grade is what the exercise is about.
If your grade came out 'metered', you are probably reading the biggest contributor rather than the weakest evidence.

05Retrieve & space

A composed figure takes the weakest grade among its parts, never the average.
Provenance is a property of each figure and has to travel with it.
A large, well-measured component does not upgrade the estimate it is added to.

06Mastery & project

Explain why averaging provenance grades rewards a bad claim, using the numbers in this bench.

Find a published efficiency claim for any system and decompose it into parts. Assign each part a grade from what the source actually says, and state the grade the whole claim is entitled to.

← What the number is per Choose the conditions, choose the answer →

PYTHON · NUMPY

The grade, not the number

Three components, three grades. Compose them correctly, then see what averaging would have claimed.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you