[B]ecause we tend to reward others when they do well and punish them when they do badly, and because there is regression to the mean, it is part of the human condition that we are statistically punished for rewarding others and rewarded for punishing them.That's Daniel Kahneman, more here. Regression to the mean will occur in situations that involve at least some luck, which is to say, almost everywhere.
Luck seems especially inevitable once we consider that spontaneous fluctuations in your brain's dynamic states (as seen in fMRI BOLD responses) can help account for trial-to-trial variability in behavior.
One study found that the amount of pain that people feel following laser stimulation (equivalent to a pinprick) can be predicted (beyond 5% chance) based on baseline, spontaneous activity in certain brain regions three seconds before the stimulation. (pubmed, PNAS)
Another study found that 74% of the within-participant variability in a button press force task could be attributed to ongoing fluctuations in neural activity. (pubmed)
In light of the fact that seemingly uncontrollable neural fluctuations play such an important role in behavior on any given attempt, punishing people for poor performance on small sample sizes seems particularly pernicious.
But then again, most punishment is probably not really intended to improve future performance, but rather to improve the mood and status of the punisher.
