Damage-formula teardown

Ten logged hits, the formula fitted to them, the predicted eleventh, and one page of method — marked on the method more than the fit

Due: 23 April 2027 · Weight: 20%

Week 3’s exercise, on a game nobody has handed you the numbers for.

What you hand in#

Three artefacts and a page.

The log. Ten hits, with every held-constant input recorded beside each one, and the constants named. A log that doesn’t say what was held still is a list of numbers.

The fit. A formula, and — stated before it — the curve shape you assumed and why. Linear, quadratic, cubic, capped.

The prediction. The eleventh hit’s value, a tolerance you committed to in advance, and the number you actually got.

The page. One side: what you varied, what you controlled, and what you would do differently.

The capability it tests#

Week 3’s: reverse-engineering a damage formula from ten logged hits. The method is weighted at 60% because the method is the part that transfers — you will use it again in week 8 on a stagger threshold and in week 10 on an experience curve, where nobody publishes the numbers either.

Marking bands#

Strong. The shape was guessed before the fit and the guess is on the page. The tolerance was committed to in advance. The prediction landed inside it — or it missed badly and the page diagnoses why from the log itself, which scores the same. Controls are named specifically enough that somebody could repeat the experiment.

Adequate. Ten clean hits, a fit that describes them, a prediction that is roughly right. The method page recounts what was done without saying what was controlled or why that shape was expected.

Weak. A good fit with no stated tolerance, which can’t be wrong and so isn’t a model. Ten hits with a stat changing between them. A shape chosen after seeing the curve and presented as a hypothesis.

When it’s due, and what happens before#

The deadline is at the top of this page, three weeks after week 3’s session, so the fitting happens while the material is close.

The saves and recordings go up at the start of week 3. Bring a half-finished fit to week 4’s session — expected values are easier to argue about with your own formula in front of you, and a fit that survives contact with the roll is usually the one that survives marking.