Null distribution

Every p-value is measured against the null distribution, which answers what your statistic would look like if the null hypothesis were correct. You get that distribution either from theory (a normal, t, or chi-square model) or from simulation, as in a randomization test. For example, under $H_0: p = 0.50$ with n = 100, the null distribution of p-hat (the sample proportion) is approximately normal, centered at 0.50 with standard deviation $\sqrt{0.50(0.50)/100} = 0.05$. An observed p-hat of 0.60 then sits two standard deviations out, giving a two-sided p-value of about 0.0455.