Overconfidence is a psychologically-generated distortion of perception consisting of an inflated estimation of one's future success. Because both economics and law usually encourage accurate and calculated estimates of risk, overconfidence is regarded to be detrimental to human judgment and it is denounced as a vice. When an assumption of risk is subsequently realized, the presence of overconfidence is thought to prove that the assumption of risk was ex ante unwarranted. The main argument of this paper is that overconfidence is underrated. Overconfidence can have constructive value in a variety of situations in which people tend to display a hyperbolic discounting of future utility, which produces in turn inconsistencies in their preferences over time. Intuitively, by artificially raising the estimation of future rewards overconfidence can be instrumental if it offsets the inhibitory effects of exaggerated preference of present rewards. Thus it can produce greater incentive, greater perseverance, resolute performance, and consequently higher achievements. Examples of this effect are found in situations of negotiation, in the improved recovery rate from serious disease and in situations of violent confrontation or wars. Because overconfidence can lead to positive outcomes, there is at times an incentive to manipulate the facts perceived by the decision maker in order to build up his confidence and increase the efficiency of the decision. These manipulations are a branch of the general theory of non-cooperative games with non-perfect information. We particularly draw attention to internal manipulations, performed by the same player between two points in time, where the former concerns the latter. These manipulations consist in constructive self-deceit and include, inter alia, selective memory, deliberate forgetfulness, and calculated inadvertence. In light of the constructive value of overconfidence, we regard these common mental practices as possibly rational and beneficial. It follows that under some circumstances, and in particular where the assumption of calculated risk should be encouraged and the chilling effect of over-deterrence should be eschewed, a systematically erroneous assessment of the risk is a key to optimal behavior. Therefore, we criticize the common supposition, according to which an overconfident person should be deemed negligent even if his overconfidence proves useful. We further object to the common view, according to which an intended manipulation of memory should necessarily entail recklessness. For similar reasons, we argue that the law of complicity should acknowledge friendship and love as countervailing the need to combat crime, inasmuch as friendship and love are instrumental to self-esteem.