Muscle coactivation can be defined as the simultaneous activation of agonist and antagonist muscles acting on a given joint. While it may seems wasteful to coactivate agonist and antagonist muscles since their reciprocal action would impair the resultant muscle torque, to use this strategy can be useful when we learn novel tasks, when we perform a movement requiring a high degree of accuracy, when we need to protect a joint. According to literature, the muscle activation patterns characterizing a given motor task should be quite constant in different subjects. This observation suggests that the CNS operates a selection among a number of pre-existing underlying solutions, somehow limiting the number of muscle activation patterns that can be activated to perform a given task. This, in turn, contrasts with the evidence that net torque exerted around a given joint can be obtained using many different activation strategies. Both these hypotheses - few muscles, greater net torque - point to the "economy" of movement as the key to decipher some puzzling choices of our motor control system and have produced conflicting results. A somewhat different point of view on coactivation, supported by results from athletes, will be presented in this lecture, focusing on how agonist-antagonist coactivation patterns can be modeled and adapted in response to specific form of athletic training where "elegance " and "versatility" will be the key words.