The three variables
Passage difficulty (common words vs technical text) moves scores ±15%; warm-up state (first test of the day runs 5–10% slower); and duration (short tests favor sprinters, longer tests reward stamina). A fair baseline fixes all three: same duration, third test of the session, averaged across a week.
The warm-up protocol
Two throwaway tests before the measured one — the fingers find their calibration in the first minutes of typing, just like a runner's first kilometer. Serious test-takers run a 30-second warm-up on easy text before any measured attempt, and the measured scores stop the mysterious variation.
What your weekly average means
Five tests averaged across a week is a truer number than any single best — it reflects the speed you actually bring to work, not your lucky run. Track the weekly average; celebrate the personal bests; and let both numbers guide the practice load.