In trading it is the luck component. Two traders with the same edge and the same rules will have different months, because the order and clustering of wins and losses differ. The average is the signal; the spread is the noise.
A concrete example: with a 55% win rate, the chance that any given sequence of five trades is all losses is 0.45 to the fifth power, about 1.8%. That sounds rare. Over a couple of hundred trades, though, a run of five losses somewhere is more likely than not, and several runs are normal.
Understanding this changes how you read results. A bad week is weak evidence that something is wrong, and a good week is weak evidence that something is right. Judging a method on short windows rewards luck and punishes patience.
Variance also sets the size of drawdowns you should expect and therefore how much you can safely risk per trade. A plan that cannot survive an ordinary losing run is not a plan with a positive expectancy but one waiting for bad luck.
The practical response is to evaluate process over short horizons and results over long ones, and to size positions for the bad run, not the average one.