●Strong / ○Lean Model edge. We blend the market price with our multi-model consensus (ESPN, FanGraphs, DRatings, Torvik, Opta), shrunk hard — only ~1/6 of a model-market disagreement is real information (measured on 19,000 games). If the blended win probability beats the price by +1% it's a Lean; +2% with 2+ models on that side is Strong. This is deliberately rare: most days the market is simply right, and a quiet board is the model telling you that. Every card shows how close it came ("Best: +0.4% — flags at +1%").

★Flag (fade + model) The stacked flag — our strongest signal. Fires when two independent things line up: the public's money leaves the book carrying 70%+ of the payout risk without moving the line (the book knowingly holding — historically the book is right), AND our model consensus independently likes the other side. This combination earned +23.6% over 103 backtested bets across 2025-26, including 13-6 in a five-week window the rules never saw. Rare (~3-4 a week, MLB) and logged to the algorithm record to prove itself live.

⚡Sharp $ Books holding ⇄ML alone Money flow. Where the public money sits and how the line reacted. Raw money % is misleading on moneylines, so we convert it to book risk — each side's money weighted by what the book would actually pay out. Most of these stay context-only (backtested on 3,000+ real games: no edge). The one exception we validated: in MLB, when one side carries 70%+ of the book's risk and the book still won't move the line, fading that side earned +10.9% across three straight seasons (7,767 games tested) — that spot lights up green and leans the pick.

#1 NHL Pick #2 NBA Pick Board rank (NBA and NHL only). The top two sides of today's slate by our daily-board ranking. This is our ranking, unvalidated — it is not a Read and not a proven edge (the NHL simulation of this ranking lost 9.1% over 791 picks; the live run so far is a small sample). It never changes a card's colour. Graded separately on the Record page as board-1 / board-2.

❄Cold-visitor edge Cold-weather spot (NFL, Nov–Jan). A warm-climate visitor (Miami, Tampa, LA, Arizona, Jacksonville…) playing outdoors at 40°F or below, 25°F+ colder than they're used to. Home teams have covered 60% of spreads in this exact spot since 2005 (151 games, profitable in every era, +15% at standard juice) — a published academic finding we re-verified on 21 fresh seasons. Books shade for wind but barely for cold. Fires ~7 times a season; forecast checked automatically.

ML or spread? Which bet to place. Our flags are moneyline signals — that's what the backtests measured. When you'd rather work the spread, 27 seasons of NFL closing prices say: double-digit favorites → take the ML (laying 10+ has cost 4–7% while the favorite ML is priced nearly fair); double-digit underdogs → take the points, never the dog ML (−13 to −34% historically). Favorites win-without-covering 23% of the time at −7 and 34% at −10+, so the number matters more than the winner. On 3 and 7 the half point is real money (−3 → −2.5 is worth paying up to about −131). The ⊕ Pick a side button prices ML vs spread for every game and names the exact bet.

≠Models split Model disagreement. Our sources disagree with each other by 8+ points — high uncertainty, and the market is usually the one that's right when models scatter. Treat gaps on these games with extra suspicion.