Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Card predictions use the league-average referee. A match's referee is rarely assigned when these predictions are made, about three days before kickoff, so the specific referee's tendencies are not part of the model.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Arrows show which of these mattered most to the model for this specific prediction, and in which direction.
Recomputed about once a minute from the live score. Separate from the locked pre-match predictions above, and never graded.
At 90' (1-2): home 0%, draw 3%, away 97%.
Kept public, never deleted. Each one says why it was voided.
Slated 44 days before kickoff, outside the 3-day slate window (ADR-012); superseded by a fresh slate.