China has experienced unprecedented economic growth and changes in urban form in the past decades. Increased urbanisation and motorisation puts pedestrians and automobiles at greater conflict. Because of China's long urban blocks (superblocks), many conflicts occur mid-block at informal or illegal crossings. This study focuses on factors influencing mid-block crossing and gap acceptance. We remotely observed illegal mid-block crossing of a six-lane urban arterial in Kunming, China, tracking 522 accepted gaps and 152 rejected gaps in a two-stage crossing (roadside to median and median to roadside). We fit a probit discrete outcome model to the data to estimate environmental determinants of gap acceptance (and rejection) behaviour, including gap size, vehicle speed, time waiting and gap lane position. We also estimate a conflict model, focusing on parameters that influence the probability of vehicle speed changes or lane deviations.