Why is it not enough for you to just monitor their performance, i.e. correlation between estimated implementation time and actual time before the task is ready?

You have 6 different formulas for calculating ‘hours’, they look similar but it’s difficult to generalize them (seems it will introduce more complexity than simplification) . 6 formulas means you have 6 implicit functions. So you need a pattern that allows easily implement variability of the function used. That is “strategy” pattern is ideal for you, initial proposal was good enough to look for ‘better’ solutions.