Determining prognosis in acute liver failure (ALF) is vital to consideration of transplantation for these critically ill patients. Transplanting a likely survivor may be as unfortunate as not transplanting someone whose outcome is very poor. Prognostic scoring systems have largely failed to accurately portray outcomes, partly because the use of transplantation does not allow the true transplant-free outcome to unfold. The King's College Hospital criteria have not proved accurate enough in predicting outcome for those with moderate illness; sensitivity is reasonable while specificity is poor. Use of APACHE II or MELD systems suffers from similar deficiencies. Use of specialized biomarkers such as Gc protein, alpha-fetoprotein, or troponin shows some promise but has not proven effective to date. Considering etiology and coma grade prior to development of scores based on routine or specialized biomarkers may be important. The rapid evolution of ALF as well as the different etiologies encountered may continue to frustrate our efforts to accurately predict outcomes in ALF.