In this paper we study why people fail at backward induction. To this end, we conducted a mobile experiment consisting of 27 games. Our sample includes 6,677 subjects from over 100 countries. Data indicates that low subjects' skills and high complexity of games are behind the violations of backward induction. The challenge lies in measuring skills and complexity. Using response times, we develop a measure of skills and a measure of complexity. We also find that improving skills has a relatively larger impact on increasing the probability of people backward inducting than lowering complexity.