Table 1.
Outcome distribution (1816-2015).
Table 2.
Justice-vote performance (three-class), un-adjusted assessment.
Table 3.
Justice-vote performance (two-class), un-adjusted assessment.
Table 4.
Case prediction performance, un-adjusted assessment.
Fig 1.
Case and justice accuracy 1816-2015 (by term).
Time series of the accuracy of our prediction model at both the case level (left pane) and justice level (right pane).
Fig 2.
For most of the Court’s history, Reversal was much less frequent than it is now. Only in recent history has Reversal become the more common outcome.
Table 5.
Justice-vote performance (three-class), baseline model assessment.
Table 6.
Justice-vote performance (two-class), baseline model assessment.
Table 7.
Case prediction performance, baseline model assessment.
Fig 3.
Case and justice accuracy compared against null models.
The first row corresponds to M = 10, the second row corresponds to M = ∞, and the third row corresponds to always guess Reverse. The left column corresponds to case accuracy, and the right column corresponds to justice accuracy. When our model outperforms the baseline, the plot is shaded green; when it fails to exceed the baseline performance, the plot is shaded red.
Fig 4.
Cumulative number of terms won versus M = 10 null model.
Fig 5.
Justice-term accuracy heatmap compared against M = 10 null model (1915 -2015).
Green cells indicate that our model outperformed the baseline for a given Justice in a given term. Pink cells indicate that our model only matched or underperformed the baseline. The deeper the color green or pink, the better or worse, respectively, our model performed relative to the M = 10 baseline.
Table 8.
Summary of statistical tests: p-value.