Check the scoreboard numbers

On paper, mark each round:

| Round | Truth | Score | Guess (≥0.5?) | Bucket | | ----- | ----- | ----- | ------------- | ------ | | 1 | 1 | 0.8 | 1 | TP | | 2 | 0 | 0.3 | 0 | TN | | 3 | 1 | 0.6 | 1 | TP | | 4 | 1 | 0.9 | 1 | TP | | 5 | 0 | 0.2 | 0 | TN |

Hand total: TP=3, FP=0, TN=2, FN=0. Accuracy = 5/5 = 1.00. That is your baseline. Code must match it — not invent a vibier story.

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

Beating the hand baseline on five practice rounds does not prove:

• the game works on thousands of real messages,
• every phone or every friend’s dataset,
• that accuracy alone is enough (always show TP/FP/TN/FN).

Write “proved” and “not proved” as two columns in your notebook. Validation without that split turns into bragging.

Always write the denominator: (tp + tn) / len(pairs) with len(pairs) = 5. A percent with no “out of what” can hide a tiny sample. Ask: if I only measured three rounds, would 100% feel the same?

Also try one negative thought experiment: report accuracy but hide FP and FN. Why is that unfair in a school spam demo or a shot-make game?

Before you move on, do this three-step ritual out loud with a partner:

1. Read the printed preds and point to each truth value.
2. Tick which bucket each pair belongs in.
3. Confirm the printed tp fp tn fn and accuracy=1.00 match your ticks.

If any step fails, stop. Do not “fix it later” by adding features. Validation is the gate that keeps later pages from inventing a softer target. Write the ritual in your notebook so page 8’s ship checklist can point back to it.