When the null hypothesis is true, the computed $F$ ratios in an analysis of variance follow a specific probability curve known as the $F$ distribution. This distribution is strictly unimodal and positively skewed, with the majority of values clustering around $1$. The precise shape of the $F$ distribution is dynamic and depends entirely on two parameters: the between-groups degrees of freedom ($df_B$) and the within-groups degrees of freedom ($df_W$). By knowing the exact shape of this distribution under the null hypothesis, researchers can accurately determine the $p$-value associated with any computed $F$ statistic.