Estimating risks for individual patients facilitates communication with patients, relatives and colleagues, and determines whether further treatment is futile. The process of estimating risks involves mathematics (i.e. scoring systems) and human experience and expertise. Understanding how risks are estimated is important because prognostication is an integral part of any medical specialty. In the USA, such treatment limitation or withdrawal decisions were made on only 7% of all intensive care unit patients but this represented 47% of all deaths on such units. In the UK, data reported by the Intensive Care National Audit and Research Centre suggest that although treatment limitation decisions are made on only 11.8% of patients, this accounts for over 50% of deaths on intensive care. Scoring systems offer a useful adjunct in identifying futility but there are important inherent weaknesses that limit their performance. This review aims to discuss some of these limitations.