We introduce extensive-form games (EFGs), which are a class of games that model sequential interaction and stochastic outcomes. We introduce the sequence form for representing strategies of an EFG, and an optimal way to measure generalized distances on EFG strategy spaces via Bregman divergences and dilated distance-generating functions. Then, we introduce the counterfactual regret minimization framework for learning in EFGs. We then introduce the idea of search in EFGs. Finally, we show numerical performance of some of the self-play approaches to solving EFGs.