Alternative reproductive tactics (ARTs) have provided valuable insights into how sexual selection and life history trade-offs can lead to variation within a sex. However, the possibility that tactics may constrain evolution through intralocus tactical conflict (IATC) is rarely considered. In addition, when IATC has been considered, the focus has often been on the genetic correlations between the ARTs, while evidence that the ARTs have different optima for associated traits and that at least one of the tactics is not at its optimum is often missing. Here, we investigate selection on three traits associated with the ARTs in the swordtail fish Xiphophorus multilineatus; body size, body shape, and the sexually selected trait for which these fishes were named, sword length (elongation of the caudal fin). All three traits are tactically dimorphic, with courter males being larger, deeper bodied and having longer swords, and the sneaker males being smaller, more fusiform and having shorter swords. Using measures of reproductive success in a wild population we calculated selection differentials, as well as linear and quadratic gradients. We demonstrated that the tactics have different optima and at least one of the tactics is not at its optimum for body size and sword length. Our results provide the first evidence of selection in the wild on the sword, an iconic trait for sexual selection. In addition, given the high probability that these traits are genetically correlated to some extent between the two tactics, our study suggests that IATC is constraining both body size and the sword from reaching their phenotypic optima. We discuss the importance of considering the role of IATC in the evolution of tactical dimorphism, how this conflict can be present despite tactical dimorphism, and how it is important to consider this conflict when explaining not only variation within a species but differences across species as well.