Fig 1.
Average change in tweet affect for candidates who won and were strongly or weakly expected to win (expected win and surprising win), and candidates who lost and were strongly or weakly expected to lose (expected loss and surprising loss) in Study 1.
This figure also displays change in tweet affect for all winners and all losers (Combined Win and Combined Loss). Error bars correspond to +/- 1 standard error.
Fig 2.
Average change in tweet affect for teams that won and were strongly or weakly expected to win (expected win and surprising win), and teams that lost and were strongly or weakly expected to lose (expected loss and surprising loss) in Study 2.
This figure also displays change in tweet affect for all winners and all losers (Combined Win and Combined Loss). Error bars correspond to +/- 1 standard error.
Fig 3.
The average change in tweet affect for each team in each game plotted against the relative final score difference for the team in the game, in Study 2.
The solid lines show predictions of the discontinuity regression performed on pooled data. As indicated by this regression, there is a discrete jump in change in tweet affect as teams surpass expectations.