Short projects

Six short projects in which students take a real decision on real data using the methods of the course. Each takes about 3–5 hours, uses one of the bundled course datasets, and ends in a one-page memo with specific numbers rather than a notebook full of output.

6 projects 3–5 h each real decisions, not exercises held-out set fixed and sealed Colab-ready starters

These are deliberately unlike the lab notebooks. A lab is guided, mirrors one chapter and ends with worked solutions. A project poses a problem someone actually has, gives no solution, and asks the student to commit to an answer and defend it — including, in several of them, the answer “this cannot be predicted well enough to act on.”

The six

#	Project	The decision the student must make	Data	Methods	After
1	Who should we call?	Hand back a ranked shortlist of 500 prospects and say how many policies it will sell	Caravan	Ch 4, 5	Lecture 7
2	Five numbers or seventeen?	Decide whether a board-readable five-variable model is defensible, and price the simplicity	College	Ch 3, 6	Lecture 8
3	Can you predict the market?	Tell a fund whether to trade on last week’s returns — with an interval	Weekly	Ch 4, 5	Lecture 7
4	A model the brand manager can read	Choose between the most accurate model and one you can explain	OJ	Ch 8	Lecture 10
5	What is it worth, and how sure are you?	Value five neighbourhoods, each with a defensible interval	Boston	Ch 3, 7	Lecture 9
6	How many managers can actually pick stocks?	Give a pension trustee one number — it may be zero	Fund	Ch 13	Lecture 12

The briefs, starters and notes live in Projects/. Each project folder holds three files: the brief (README.md), a Colab-ready starter notebook, and SOLUTION_NOTES.md — what a good answer finds, the trap, and a marking guide.

Open a starter in Colab — nothing to install, the data resolves itself:

1	2	3	4	5	6

Read the solution notes afterwards

The notes ship alongside the brief, exactly as the practice-exam solutions ship alongside the paper: the discipline of attempting the work first is the student’s to keep. They are genuinely worth reading after an attempt — several of these projects have a result that surprises people.

What makes them projects rather than exercises

The test set is fixed and off-limits.

Each starter defines the held-out split so that every student’s reported number is comparable, and so that Chapter 5’s honest-evaluation discipline is enforced by the scaffolding rather than by trust. Where the data are time-ordered — projects 3 and 6 — the split is chronological rather than random, because randomly splitting a time series trains on the future to predict the past and answers an easier question than the one asked.

There is a baseline to beat, computed for you.

Random targeting, the majority class, a mean-only model. A student who cannot beat it has learned something real, and the briefs say so.

Every number needs an interval or a caveat.

A point estimate without its uncertainty is not an answer to any of these questions — least of all project 3, where the interval is the whole point.

The obvious approach is often the wrong one.

Each project has a trap, named in its solution notes but not in its brief. Accuracy is the wrong metric when 6% of customers buy; random cross-validation is invalid on time-ordered returns; a confidence interval for a mean is not a prediction interval for one house; and of 289 “significant” fund managers out of 2,000, about a hundred are what chance alone produces and roughly half are significantly bad.

Where they fit in the semester

The module is graded by one written exam, so the projects are formative — none counts towards a mark unless the instructor chooses to use one. They are keyed to the lecture after which each becomes doable, in the same “one at a time, when the material is taught” rhythm as the five 60-minute short exams, and they fill the gap those papers leave: a short exam tests whether you can answer a question, a project asks whether you can decide something.

Projects 3 and 6 both end in a judgement call worth arguing about, which makes them the two best candidates for group work or a seminar discussion.

Where to go next

• Lab notebooks — the guided labs these build on.

• Advanced modules — several projects name one as a natural extension: Shapley values for explaining a targeting model, conformal prediction for the valuation intervals.

• Mock exams — the practice papers, and the exam that actually counts.

• For students — workload, and how the course is assessed.