The probability of incorrectly rejecting H0
(i.e., finding a
significant result when in truth there is no effect) increases as
the number of analyses increases.
Incorrectly rejecting H0 = Type 1 error
In other words:
When performing multiple analyses, the probability of
concluding that there is a significant difference (using your
sample), while in reality there is no difference in the
population, increases.
Compare with rolling a dice multiple times (chance on say 6).

n. a process of risk-taking in business or personal affairs wherein the causality of an event is inferred from and based on the factor of chance alone. For instance, one capitalizes on chance by selec [..]

The process where causality is inferred that is really only a random act of chance.