Track Record
The scoreboard we grade ourselves on: how often the market, the consensus of outside projection sources, and the crowd ranked the right player higher, measured against what actually happened each week. Wins and misses both stay on the board.
No grades yet.
Grades land after the first week's games are final: each source's pre-kickoff rankings get scored against the actual results, and the season scoreboard starts here. Until then the fairest score is no score.
How this is graded
Every graded week, we take the players at each position who actually played and were startable-relevant before kickoff, and form every head-to-head pair within a position. For each pair we ask one question of each source: did it rank the player who ended up scoring more ahead of the other? The score you see is the share of pairs each source called correctly, with the number of graded pairs in parentheses.
Rankings are strictly pre-game. Each week is graded against the last archived projection snapshot taken at or before that week's first kickoff, so a source never gets credit for a number it published after the games started. Weeks with no pre-game snapshot are skipped rather than graded late.
Two noise filters keep coin flips out of the score. A pair the actual results decided by under 3 points is not graded at all: that close, the outcome decided nothing. And a source is only graded on pairs it separated by at least 1 projected point; when it had the two players essentially tied, it made no call, so it gets neither credit nor blame.
Honest limits: sources are graded on different pair counts, because each is only scored where it published a number and made a call, so small samples early in the season swing hard. One week is weather, game script, and luck; a season of pairs is signal. Nothing here is start/sit advice by itself: these are projections and their receipts, not picks. Misses are published the same as hits.
Market = prices from live prediction markets and sportsbook player props. Consensus = the consensus of outside projection sources. Crowd = this site's head-to-head votes, converted to ratings. Accuracy is graded in full-PPR points.