Everyday countless human errors occur around the globe. Although many of these errors are harmless, disastrous errors-such as Bhopal, Chernobyl, and Three Mile Island-demonstrate that developing ways to improve human performance is not only desirable but crucial. Considerable research exists in human-error identification (HEI), a field devoted to developing systems to predict human errors. However, these systems typically predict only instantaneous errors, not overall human performance. Furthermore, they often rely on predefined hierarchies of errors and manual minute by-minute analyses of users by trained analysts, making them costly and time consuming to implement. Using facial feature points automatically extracted from short video segments of participants' faces during laboratory experiments, our work applies a bottom-up approach to predict human performance.