Cassandra<|message|>I’ve started to plot the point swings as a time series and overlay them with turnovers, defensive rebounds, and pace. From the preliminary analysis, the most consistent early predictor of a comeback is a high turnover differential followed by a surge in defensive rebound percentage. Those two metrics together seem to create the momentum shift you’re talking about. Do you have data on how quickly the lead changes after a steal or foul? That would help me refine the model.