4. Case Study โ€” Recovery

P/T and highest-P recover equally often โ€” both beat shortest-T and random

OrderingRecovered1st tryNot recoveredWrong procsMedian TTRMean TTR, capped
P/T (priority)282111414.3 min19.6 min
Highest P first28261515.3 min17.6 min
Shortest T first251444516.1 min27.0 min
Random (mean of 5 seeds)23.812.65.241.222.3 min32.6 min
P/T
21
7
1
2
Highest P
26
2
1
2
Shortest T
14
11
4
2
Random
13
8
8
2
1st procedureafter wrong procedure(s)not recoveredno action

TTR = injection โ†’ correct procedure(s) done. Capped mean counts unrecovered runs as the full hour.

First pick waits 60 s after the first nonzero evaluation (timed gate). P/T is not always faster than highest P: KG scores are not calibrated probabilities, and shared procedures blur the ranking.