How to read it: "fair 57.4%" means the de-vigged market median gives that side a 57.4% chance at the app's exact line. Pick'em breakevens run roughly 54.6% (6-pick power) to 58% (2-pick), so anything above your format's breakeven is a market-approved lean. Lines tagged Alt payout carry a non-standard multiplier on their app, so weigh the fair chance against that payout. PrizePicks Demons (harder line, bigger payout) and Goblins (easier line, smaller payout) are More-only picks with a non-standard payout PrizePicks does not publish. Because we can't read that multiplier, these are tagged “Payout varies · not EV-graded” and we show only the straight chance the More hits. Weigh it against the boosted or reduced payout your app shows. When the app's line differs from the sportsbook consensus, we flag it, because that difference is usually where the edge comes from. Grading uses every book quoting both sides of the app's exact line, sportsbooks and the exchanges (ProphetX, Novig, Kalshi) alike: 3+ anchors is the full-confidence band, while cards graded off just 1 or 2 anchors carry a muted "1-book anchor" / "2-book anchor" note: a real market read, but a thinner one, so weigh it accordingly. Thin-anchor lines never join the auto-built Top-EV combos. Anything above a 65% fair is treated as a data error and hidden. Lines no book quotes at all (fantasy score, mostly) carry a blue Modeled tag instead: the win % there is our estimate, built from the gap between the apps' own lines (or the player's component stat markets), capped at 62% and never treated as a market grade. Modeled picks price in your entry slip but never join the Top-EV combos, and a single-app line with no gap shows as a 50/50 Coin-Flip Line. Same de-vig method as our +EV board and public record.