Extended Data Fig. 4: Unsigned surprise over time across a game.

(a, b) We plotted the mean (a) and standard deviation of unsigned surprise (b) against game time using our metric. Mean surprise had momentary troughs and peaks at the beginning and end of each game, likely due to the momentary decrease and increase in shots taken within these short time frames. Interestingly, mean surprise tended to decrease throughout the game, owing to the probability that the outcome of many basketball games would be nearly decided at that point. However, standard deviation increased exponentially until the end of games, due to the combination of games with a near-certain outcome, which have almost no surprise, and close games, which involve increasingly rapid swings in win probability as one nears the finish.