Manual ingest uses the selected Ingest leagues checkboxes in Automation Scheduler below.

Ingest extra markets calls the per-event Odds API endpoint (one request per click; uses your API quota). Run it before scanning so alternates, period 3-way keys (h2h_3_way_h1, h2h_3_way_h2), exchange h2h_lay, spreads_h1, totals_h1, alts, and DNB exist in the database. Ranking: fragile extras (alt spreads/totals including H1 alts, team totals) need a best price ≥ your Min best decimal (fragile extras) (default 1.45; same setting lives under Automation → Market filters) and simple EV (model prob × best odds − 1) ≥ 2%; only the top edge is kept per team spread side (full game and H1 separately) and one Over / one Under per game (and per half for H1) so ladders do not flood the card. The model compares all ingested book quotes (min / avg / best, range %, # books) and can add a small edge boost when several books disagree but the best price still passes the consensus outlier check. Same-game cross prices (below the table after a scan) look for two- or three-way combinations on this game where the best price on each leg (possibly different books) implies under ~100%—including paired totals (Over/Under same line), complementary spreads (home vs away, lines sum to ~0 at the same handicap size), and ML / 3-way sums—separate from the per-line +EV rank. Positive-outcome note does not change the order of lines; it only adds text that the shortlist can mix market types if those score best—each row is still its own edge, not a parlay. Stake units apply money management: shorter best decimal and extra-market type trim the tier stake (shown as (tier x) when reduced). AI note is from the same on-site edge model as auto-picks—not a separate chat LLM.