Ship checklist for your mini prediction game

Your ship package is small on purpose:

• the working path,
• the expected evidence line,
• proved vs unproved,
• privacy + honesty gates,
• how to reproduce a rules bug.

Here is the artifact you can still run live:

truth = [1, 0, 1, 1, 0]
scores = [0.8, 0.3, 0.6, 0.9, 0.2]
threshold = 0.5
preds = [int(score >= threshold) for score in scores]

pairs = list(zip(truth, preds))
tp = pairs.count((1, 1)); fp = pairs.count((0, 1))
tn = pairs.count((0, 0)); fn = pairs.count((1, 0))
accuracy = (tp + tn) / len(pairs)
print(preds, f"accuracy={accuracy:.2f}", tp, fp, tn, fn)

Expected evidence:

[1, 0, 1, 1, 0] accuracy=1.00 3 0 2 0

1. Happy path — print matches the expected line on two runs.
2. Validation — TP+FP+TN+FN = 5; accuracy matches hand count.
3. Failure handling — empty lists and mismatched lengths fail loudly.
4. Observability — a friend can read threshold, preds, and buckets from the output.
5. Safe demo rules — no real classmate messages; don’t claim the toy works on all phones.
6. Proved vs unproved — proved: exact scoreboard on five practice rounds; unproved: real-world spam/shot performance everywhere.

Limits are part of the craft. Saying “n=5 fixture only” is stronger than a vague “it’s just a demo lol,” because it tells the next person what evidence exists. School projects that write proved/unproved earn more trust than projects that only show a perfect accuracy sticker.

Shipping here means: the notebook runs, the evidence is pasteable, the safe demo rules are visible, and a partner can reproduce both the healthy path and one game bug. It does not mean uploading to every phone, connecting to a real school inbox, or promising accuracy outside the fixture.

When you present, lead with the decision the toy supports (“choose a cutoff and see error buckets”), then show the printout, then read the unproved line. That order stops the audience from hearing only “accuracy=1.00” and walking away with the wrong story. Keep the four kid metaphors handy: TP = correct yes, FP = false alarm, TN = correct no, FN = missed yes.