We test our candidate drivers in our offseason competitions. By the last competition we have a few pairs of drivers to choose from. Students are evaluated on their general maturity, response under pressure, ability to follow instructions, and
then raw driving ability. Students, even those who are introverted and don't want to speak up, are encouraged to drive during quals no matter what their skill level is. We see the broken robots as an opportunity to train the build team.
We then plan for what training and level of practice is needed in the upcoming season. Student pairs are not broken up at that point - if one driver doesn't work out, neither does the other. We have had a couple of years where a single student is part of multiple pairs, like this year. Before the first event we have narrowed down the drivers to two sets, choosing the best one for the first event. Yet final selections are never 'final' until elims at the first official event.
We have even had some week 6 surprises in recent years. In 2015 we tossed the entire plan in favor of adapting to what our students students proved they were capable of at the event, which turned out to be quite different than both the offseason events and the build season.