Sports, due to their global reach and impact-rich prediction tasks, are an exciting domain to deploy machine learning models. However, data from conventional sports is often unsuitable for research use due to its size, veracity, and accessibility. To address these issues, we turn to esports, a growing domain that encompasses video games played in a capacity similar to conventional sports. Since esports data is acquired through server logs rather than peripheral sensors, esports provides a unique opportunity to obtain a massive collection of clean and detailed spatiotemporal data, similar to those collected in conventional sports. To parse esports data, we develop awpy, an open-source esports game log parsing library that can extract player trajectories and actions from game logs. Using awpy, we parse 8.6m actions, 7.9m game frames, and 417k trajectories from 1,558 game logs from professional Counter-Strike tournaments to create the Esports Trajectory and Actions (ESTA) dataset. ESTA is one of the largest and most granular publicly available sports data sets to date. We use ESTA to develop benchmarks for win prediction using player-specific information. The ESTA data is available at https://github.com/pnxenopoulos/esta and awpy is made public through PyPI.