This article empirically investigates the effects of guilty pleas on sentencing in Brazil, with particular attention to the constraints imposed by the Supreme Court’s Theme 158 and the Superior Court of Justice’s Binding Precedent 231. Using quantitative methods and an original dataset of criminal judgments for theft (simple and aggravated), robbery, and drug trafficking, the study applies multiple linear regression models to estimate the impact of a guilty plea on custodial sentences. The results show that confession is frequent behavior (present in 49% to 71% of defendants) and leads to substantial sentence reductions (between 64 and 216 days, depending on the offence). However, in a significant share of cases these effects are neutralized by the constraints imposed by the higher courts, producing the paradox that defendants who plea guilty often receive no benefit at all. The empirical findings reveal measurable harm stemming from the prohibition on reducing sentences below the statutory minimum and reinforce the need for a critical reassessment of the current jurisprudence, in order to ensure the effectiveness of the constitutional principle of individualized sentencing and the fundamental right to a just punishment, especially in a country with alarming levels of imprisonment and degrading conditions of life in prisons.