Dealing with regression to the mean

If I have two time points T1 and T2 and separate the groups based on their performance at T1 (High v Low), then after treatment assess their performance at T2, how do I deal with regression to the mean?
What is your goal? Why do you separate the groups based on performance?
It appears the "treatment" is only working for those with "impaired performance" at baseline. The treatment improves their performance but has no effect in those already performing well at baseline. I know I could do another study with a more sophisticated design but was just interested to know if there was a way to remove regression to the mean issues.


Ambassador to the humans
Why was the treatment assigned in this manner in the first place as opposed to randomly assigning units to the groups?