I'm currently trying to perform a one-way ANOVA test on three groups of race bike athletes following a different training schedule each. The athletes were randomly selected and arbitrarily divided into three groups (20 athletes each). After the training period every athlete's FTP is measured (maximum power sustainable for the duration of one hour). Based on that data I will perform the one-way ANOVA test to tell, if there are significantly different results.
I know that before applying the test, I have to assure that the data fits some requirements like having an approximately normal distribution.
Here's my question:
Do I have to test each group for normality with the FTP results after the training schedule or do I have to check the sample data for normality before the groups are selected and the training started?