The scorecard (/scorecard, linked from the footer) grades the Seer against what happened, on every league the app has loaded, not just yours. It is updated daily and shows the sample behind every number.
Playoff odds: every day the app stores each team’s playoff and title odds in every in-season league it knows; once a season finishes it checks who made it. The calibration table groups the odds it gave (0–10%, 10–25% and so on) and shows how often those teams really got in. A well-calibrated model’s “50–75%” teams make it about 60% of the time. The Brier score is the error of the probability itself; the “always the base rate” score beside it is what a model that knew nothing would get.
Trade verdicts: every trade the app has seen since the daily values began is priced on the trade day (the “won / lost / even” the Trade Calculator would have shown) and again on today’s values. The table shows how often the verdict held eight weeks or more later, and how often a “won” turned into a loss.
Projections: the Lineup page’s start/sit odds rest on Sleeper’s pre-game projections. The scorecard compares them with the box score for every QB, RB, WR and TE week stored this season, in full-point PPR as a common yardstick: the average miss, the bias, and by projected bucket how often a player at least hit his number.
The Seer grade: the prospect model’s walk-forward accuracy on every draft class since 2008 (the same numbers as the Profile lab), with its calibration.
Until the database has a finished season of odds behind it, the playoff section says so rather than showing a thin sample.