# Marble Probability Lab — analytics map

## Outcome and interaction

Learners connect a post-change probability to an algebraic model, then update favourable and total counts for two draws without replacement. The shared model supports deliberate trials of `n`, an explicit “add 9” transition, and a visible first-red removal.

## Evidence map

| Evidence | Interpretation | Student use | Teacher use |
|---|---|---|---|
| Submitted answer and confidence | Result and certainty at check time | Compare confidence with evidence | Identify confident misconceptions |
| Trials of `n` and “add 9” actions | How the bag was modelled | Review a systematic or guess-and-check path | Distinguish modelling from a bare answer |
| Factor selections and red-removal action | Whether dependence was represented | See why both counts change | Diagnose replacement errors |
| Hints, attempts, revisions | Scaffolding and change after feedback | Notice productive revision | Target follow-up support |
| Completion score | Final checked outcome | Monitor completion | Record 0–2 marks |

## Stable misconception codes

- `uses-added-red-as-original`: 9 reported as the original red count.
- `confuses-original-final-red`: 21 reported as `n`.
- `reports-final-total`: 33 reported instead of `n`.
- `algebra-error`: another numerical value after a valid submission.
- `same-probability-twice`: uses 21/33 for both draws.
- `red-count-not-reduced`: uses 21/32 on draw two.
- `total-not-reduced`: uses 20/33 on draw two.
- `adds-sequential-probabilities`: adds rather than multiplies the two draw probabilities.
- `second-draw-only`: gives 20/32 alone.
- `invalid-response`: response cannot be evaluated safely.

## Transport and privacy

Micro-events remain local until completion. On pause or exit, an explicitly in-progress state is saved only when meaningful evidence exists. The final payload contains two stable item records, bounded semantic history, a visual teacher report, and a diagnostic summary. Raw keystrokes, authentication parameters, learner identity, and high-frequency pointer events are not stored.

## Teacher review checklist

- Confirm the wording and 1-mark allocation for both parts.
- Test keyboard-only use, touch targets, 320px reflow, and the 450px first viewport.
- Launch in the target SLS lesson and confirm restore, final state, score `0–2`, and rich feedback.
- Confirm that the activity does not submit an untouched launch.