League — All / NFL / CFB. Switching this automatically jumps Week to that league's earliest tracked week (or back to "All" if you pick "All League").
Week — follows League automatically, but you can always override it manually, including picking "All" yourself at any time.
Conference and Search — narrow the list further; these don't affect League/Week's auto-behavior.
Season-long, top of the page: SU (straight-up), ATS (against the spread), O/U (totals), and 🏆 GP (Golden Picks' own record). These only count games that have actually finished and been graded — an ungraded week shows 0-0, not a guess.
The small line underneath says "✓ server-verified record" when the number comes from the server — graded server-side and shared across every device, not just this browser — or "local record (unverified)" when it's still falling back to this browser's own tally (e.g. right after a game finishes, before the server has caught up). The server record is the one to trust; the local fallback exists only so the ticker is never blank.
The site's own top-10 ATS picks for the week. Active every Wednesday through Monday night — outside that window the button is greyed out and tells you when it reopens. Click it any time to see the current list, or check why it's not available yet.
Ranked by a score, not just the displayed edge% — raw Kalshi market edge (how far the cover probability sits from a coinflip), plus a bonus if the line has moved 3+ points (sharp money, +15 to the score), times a discount on blowouts (any game with a spread bigger than 21 points has its score cut to 70%, since a lopsided score doesn't necessarily mean a lopsided cover). This means a lower displayed edge% can rank above a higher one when a bonus or discount applied — when that happens, a small note under the edge% says exactly which one and why, so the order is never a mystery.
Below the matchup: for CFB, the small badge next to each team is their AP poll rank (or "NR" if unranked). For NFL, it's their current season record. Team names show season record either way. Weather, when it's meaningful, appears in this line too.
Team name colors: green = the market favorite (the team the spread favors), white = the underdog.
Shows the opening line (struck through, if it's moved) and the current line. Underneath is a note explaining where the pick actually comes from — it'll say one of two things:
"Kalshi: favorite covers X% of the time" — this is the strong, preferred signal. Kalshi is a real prediction market where actual people trade real money on whether the favorite covers this exact spread. This percentage is read directly off that market, not a model's guess — it's what the crowd is actually paying to bet on right now.
"Model gap: X pts vs. posted line" — this is the fallback, used only when Kalshi doesn't have a tradeable market close enough to the actual posted spread to trust. It's a synthetic estimate built from team win-probability math instead of real market pricing — a reasonable backup, but weaker evidence than a real market number.
If weather is severe enough to matter, you'll also see a caution or an adjustment note explaining exactly what changed and why.
O/U always shows a real OVER or UNDER pick whenever Kalshi has a total market for the game — even a near-coinflip read still resolves to a direction (tagged Toss-Up tier, shown dimmed) rather than being withheld. The only time you'll see "TBD" is when there's genuinely no market number to go on at all — no Kalshi total AND no posted sportsbook line yet, which happens on smaller/lower-interest games before books get around to pricing them. That resolves on its own once a line posts; there's currently no second, independent total-model to fall back to before then.
Real market prices (Kalshi) are the dominant signal behind every pick — normally 100% of it. But when independent statistical models (efficiency ratings pulled weekly from ESPN FPI / SP+ / Sagarin / Massey, plus same-day injury-driven updates) disagree with the market by 3 or more points on who covers, that's a genuine, real disagreement worth weighing — not noise. When that happens, you'll see a line like:
"📊 Blended (72% Kalshi): 44%→38% — SP+ discrepancy (4.2 pts, injury adj)"
Read left to right: Kalshi still gets the majority weight — 72% here — scaled by how convicted Kalshi's own price is (a closer-to-50/50 Kalshi price yields more room for the stats disagreement to matter; a lopsided one barely moves). The two percentages are the cover probability before → after the blend. The source (here, SP+) and the size of the disagreement (4.2 pts) are named so you can judge it yourself, not take it on faith. No disagreement, or one under 3 points, means no blend note at all — the pick stays 100% raw Kalshi, and the "Kalshi: favorite covers X%" line above it is the whole story.
This is real, live-computed logic — not a placeholder — but it's honest about its own limits: efficiency ratings only exist once games have actually been played (there's no legitimate preseason number), so you won't see this note before then, and it only ever nudges the pick, it never overrides a real market price outright.
Each comes with a confidence tier: Lock (high conviction), Lean (real edge, more uncertainty), or Toss-Up (essentially a coinflip — shown for completeness, not as a strong recommendation).
Got a different number from your own sportsbook? Enter it here to see how the pick would change against your actual line, without affecting the site's own displayed pick.
Appears only on cards for games currently in progress. Use it to check whether an in-game (live) bet is worth making — it does not place a bet or pull odds from anywhere; you type in the live spread you're seeing at your own sportsbook right now, and it does the math.
Enter something like "TCU -3.5" and hit Check. Two important things about how this works:
It grades on the FINAL score, not the score at the moment you bet. A live spread bet is still a full-game bet — if you take TCU -3.5 while TCU leads by 10, TCU still has to win the actual game by more than 3.5 points at the final buzzer for that bet to win. Building a big lead and then coasting to a narrow win can still lose a live spread bet. This tool's cover math reflects that correctly: it shows the current margin, the net points still needed by final, and (when clock data is available) a pace-based projection of whether that's realistic given time remaining and each team's scoring rate so far.
Every number is shown, nothing is a black box. The verdict (Lean / Toss-Up / Fade) is plain arithmetic — current margin, points needed to cover, minutes remaining, and each team's share of scoring so far — not a hidden model. If there's no clock data for a game (this can happen when the backup score source is being used instead of the primary one), it says so and gives you score-only math instead of quietly guessing at pace.
Once you enter a line, live team stats (rushing/passing yards, turnovers, time of possession, 3rd-down efficiency) fetch automatically and appear below the math, along with a short trend read on turnover margin and possession share. Your own pre-game injury notes for both teams also show here for context — this tool can't detect a fresh in-game injury (that's not in any stats feed reliably), only remind you of what was already known going in.
Expand any card for a five-part breakdown: Matchup Overview, Analytics Profile, Betting Market context, Situational Factors, and an Injury Report. This is written context to inform your own judgment — it doesn't change the numeric pick above it.
Dark or light — top right, saved automatically for next time you visit.
A behind-the-scenes readout of how many of this week's ungraded games actually have each data source resolved right now (Kalshi, weather, injuries, analysis, model, and remaining Odds API quota for the month). Useful for confirming the site itself is healthy, not for picking games — hidden from the main view by default so it doesn't clutter the page.