This article offers a unified theoretical framework to address two distinctive forms of adversarial procedure: the bona  de adversarial system and the pseudo-adversarial system. In the former, a harsh contest between the prosecution and the defense is promoted, and an acquittal is rendered with sufficient likelihood. In the latter, the prosecution overpowers the defense so that defendants are almost always convicted. We explain this institutional divergence as a result of optimal incentive designs adapted to diverse environments. Policy variables of these incentive designs include a judge's standard of proof beyond reasonable doubt, a prosecutor's discretionary rule for indictment, and a defendant's right to counsel. Our theory suggests that the bona  de adversarial system functions best with jury trials, publicly-elected prosecutors, and competent defense counsels, while the pseudo-adversarial system functions best with bench trials, bureaucratic prosecutorial offices, and compromised counsels. We further investigate how these two adversarial systems can simultaneously persist.