Movement
Every figure here is summed from the matchup log itself — the per-vote rating changes that were actually applied — not inferred by differencing two daily snapshots. Quarantined and discarded votes are excluded.
No rating gained in the last 7 days.
No rating lost in the last 7 days.
Ranked by the Glicko-2 expected score computed from the state at the moment of the vote — both ratings and both deviations, as recorded on the matchup row. Votes cast before those deviations were recorded are excluded rather than scored against a substituted default, because an assumed deviation would manufacture upsets that never happened.
No result so far has gone against the ratings. With this few votes almost every pairing is a genuine coin-flip, so there is nothing yet for the model to have been surprised by.
Pairings closest to an even split — the debates the crowd genuinely cannot settle.
No pairing has reached 50 votes yet — the most-voted has 1. The list below is the real leader board at today’s volume, shown so the page is not blank, but it does not yet clear the bar this board is meant to apply.
The same disagreement, weighted by turnout: votes × p × (1−p). A 55/45 across four hundred votes outranks a 50/50 across four, which is why this is not simply the list on the left.
No pairing has reached 50 votes yet — the most-voted has 1. The list below is the real leader board at today’s volume, shown so the page is not blank, but it does not yet clear the bar this board is meant to apply.