Every time when a dispute arises if the pretest were just bad or the worst in the history, I wondered what is a typical part of submissions failed on full tests. I ran my script to calculate this value for Codeforces rounds. The value is based on 895 Div.1 and Div.2 rounds which were held since 2012. The histogram for 4846 tasks is shown below
.
Let's draw a boxplot for this data
But it is really more interesting to look at distribution of failed submissions by task rating
The same data in the number format
If you still feel angry, guilty or depressed about a failed submission, I could recommend a cool book for CP guys — analytical-minded author describes problems by categories and their solving algorithms )