Counting what moved
One nudge already moved the rule. Only a new count can say what the nudge did.
- Start: The rule missed 2 of the 4 kept-back days.
- First nudge, on the heavy-use day: The rule missed 1 of the 4 kept-back days.
- Second nudge, on the hot-week day: The rule missed 0 of the 4 kept-back days.
The rule
Guess, compare, count, nudge, again. A count over a few days can hide what a nudge broke.
Count before the nudge, and count again after it.
After nudge one, the weights are 12 and 4. The rule misses 1 of the 4 kept-back days. The rule now also gets two other days wrong. Those are gauge 4, days 2 and gauge 3, days 1. Take gauge 4, days 2. Gauge: 4 × 12 = 48. Days: 2 × 4 = 8. The sum is 56. Take gauge 3, days 1. Gauge: 3 × 12 = 36. Days: 1 × 4 = 4. The sum is 40. Both sums call, but the caretaker waited on both days. So count all 12 days too. The first nudge takes whole-log misses from 2 to 3.
Run the loop a second time. The day is gauge 5, days 3, a hot-week day. The caretaker waited. Its sum is 5 × 12 + 3 × 4 = 72, too high. So this nudge moves both weights down. Gauge weight: 12 − 5 = 7. Days weight: 4 − 3 = 1. The kept-back miss count reaches zero. The whole log of 12 now shows 1 miss.
Measuring only the day just nudged would say 0 of 1. That count looks too kind, since the nudge fitted that very day.
One day still cannot come right. Early in the month the caretaker called on gauge 5, days 3. On the hot-week day she waited with the very same two numbers. Two weights and a threshold give both days the same sum. So the rule must miss one of them.
Watch out!
Missing 0 of 4 days looks kinder than it is. Two of those four days were used for nudging. A fair test needs days no nudge has seen.
Tap the five cards in loop order, twice through. Find the count that gets worse.
Loading lab…