Sleep deprivation is strongly associated with worse psychomotor vigilance, including slower responses and more lapses. Even without total sleep deprivation, feeling tired or distracted can create a long tail of unusually slow trials.
Attention also fluctuates naturally. That is one reason a set of 10 trials usually tells you more than a single fastest click.
Reaction time changes across adulthood on average, although there is wide overlap between individuals. Large studies show age-related slowing, while physical health, cognition, motor speed and experience also contribute.
See the dedicated reaction time by age guide for a more careful interpretation.
Repeated trials can improve a score because you learn the rhythm, become comfortable with the input and reduce uncertainty. Practice effects are real, especially early in repeated testing.
But anticipation can also produce false starts. Randomized waiting periods are useful because they make it harder to simply predict when the go signal will appear.
A recent meta-analysis of randomized caffeine studies found small average improvements in reaction time and accuracy on attention tasks in rested adults. Meta-analyses of acute exercise also report small cognitive benefits in some conditions.
Neither finding means you should chase browser scores with stimulants or intense exercise. Dose, timing, health, habituation and task all matter, and a game score is not a health outcome.
For personal tracking, the best way to reduce these confounds is boring but effective: keep your task, browser and device as consistent as practical.
Fatigue, attention, practice, stress, setup changes and normal trial-to-trial variability can all move the result.
On average, reaction time changes with age, but age does not determine an individual score. There is substantial person-to-person variation.
Yes. Display, touchscreen, mouse, keyboard and system timing can all add latency to the recorded browser result.