Methodology

Betting 101 — the markets on every pick card

Every pick on this site is tagged with one of six market types. If you don't bet, here's the two-minute version — with each market's actual record across the corpus, because the honest lesson of 20,000 tracked picks is that none of these are money printers.

Spread

A bet on the margin. "Chiefs −3.5" wins only if the Chiefs win by 4 or more; "Jets +7.5" wins if the Jets lose by 7 or less (or win). The number exists to make both sides a coin flip. Corpus record: 5928-5739-224 (50.8% cash rate, -2.7% ROI) — better than a coin flip, and still underwater, because of the vig (explained in the Lab).

Total (over/under)

A bet on the combined score, either side of a number: "Over 44.5" needs 45+ total points. Corpus record: 2305-2162-98 — as close to a literal coin flip as 1,100+ picks can get, and it still runs -1.9% ROI. That gap between "won half" and "made money" is the vig, the house's cut baked into every price.

Moneyline

A bet on who wins, no margin involved. The price does the work: "Norway +300" pays 3 units per 1 risked because Norway probably loses; a −300 favorite risks 3 to win 1. Corpus record: 2557-2804-69 — losing more often than winning, yet -4.9% ROI, because underdog wins pay more than favorite losses cost. Moneylines are why cash rate alone can't rank anyone (our leaderboard ranks by units for the same reason).

Prop

A bet on a player's stat line, not the game: "Mitchell over 27.5 points." Props are priced in thinner markets with wider vig than game lines, and the corpus shows it: 2675-4115-12, -8.4% ROI.

Parlay

Several bets chained into one ticket: the payout multiplies, and every leg must win. One loss kills the ticket — which is why books advertise them so cheerfully. The corpus record is the clearest anti-endorsement on this site: 491-1066-1 (31.5% cash rate), -18.6% ROI. Long odds aren't the same as good odds.

Futures

A season-long outcome settled months later: championship winners, win totals, awards. Your money is locked up the whole time, and the corpus runs 1261-4043-10 (23.8% cash rate, -6.1% ROI) — most "who wins it all" calls simply lose.

The vocabulary on a pick card

Odds are quoted American-style: −110 means risk 1.10 to win 1; +140 means risk 1 to win 1.40. Line is the number attached to a spread, total, or prop. Units is profit measured in bet-sizes — every pick here is graded as 1 unit at the quoted odds (−110 when none was stated). A push lands exactly on the line and refunds; a void is a bet that couldn't be graded fairly and counts for nothing. The green CLV chips compare a pick's number to where the market closed — the Lab explains why that's the measure sharps care about.

How picks are captured

We pull each show's published episode audio, transcribe it, and use a language model to find genuine betting recommendations — the segments where a host actually says "here's my play." Jokes, hypotheticals, and recaps of someone else's pick are excluded.

Every pick has to survive scrutiny before it counts — most of it automated, some of it human. Each pick carries the verbatim quote it came from, and it only publishes after passing checks that the quote actually exists in the transcript at the stated moment, that the selection, line, and odds follow from what was said, and that a second, independent model pass agrees the host genuinely committed to the bet rather than musing about it. Picks the system isn't confident about are held back until a person reviews them. A human has audited thousands of picks across repeated full-corpus sweeps — and the whole corpus has been re-checked against the transcripts more than once — but with picks numbering in the tens of thousands, we won't pretend human eyes have read every single one. If you find one that doesn't hold up against its own receipt, that's a bug, and we want to know.

The ▶ button on a pick card plays the episode at the moment of the pick, streamed from the show's own public feed — we never host their audio. One honest caveat: podcast feeds insert ads server-side, and the ad load changes between listens, so the drop-in point can drift a little from the exact sentence. The player's ±15s buttons exist for precisely that hunt. The quote and timestamp themselves come from the transcript each pick was graded against, so they stay accurate even when the feed's ad load shifts.

How picks are graded

• Spread — the team's final margin is compared to the line. Land exactly on the number and it's a push.
• Total — combined final score vs. the over/under. Exact number pushes.
• Moneyline — did the picked team win the game.
• Props, futures, parlays — graded by hand for now; they're flagged rather than auto-settled.

Final scores come from ESPN. A pick stays pending until its game is final. A disclosed wager whose recorded terms can never establish an outcome is retained only in the internal audit as ungraded; it is not published or included in any site statistic.

Units & ROI

We assume a flat 1 unit staked on every pick. A win pays out at the odds quoted on air (e.g. +150 returns 1.5 units; −110 returns 0.91); a loss is −1 unit; pushes and voids are 0. When a host didn't state odds, we assume standard −110 juice. ROI is net units divided by the number of settled picks. Split-stake Asian handicaps can produce a half-win or half-loss: the record follows the net result, while units credit or debit only the half of the stake that did not push.

Closing line value

For game markets (spreads, totals, moneylines) and player props we also capture the closing line — where the betting market finished just before kickoff. Historical picks come from archived odds snapshots (The Odds API); new picks are captured live in the final minutes before each day’s slates close. A pick beat the close when the market moved toward it after it aired: take −2.5 and watch it close −4, and the market ended up agreeing with the pick. Consistently beating the close is the standard test of real forecasting skill, far less noisy than short-term win/loss.

• Which book: the book named on air when our odds source carries it, else the show's house book for that era, else the median across US books ("consensus"). ESPN BET and Hard Rock Bet aren't carried by our source's primary US feed, so historical picks quoting them use the consensus close — each pick's tooltip shows which was actually used. New picks (from July 2026) are captured with the secondary feed included, so those books' own closes are recorded going forward. Re-measuring everything against single books (DraftKings, FanDuel) shifts beat-close rates by well under a point, so the tiering doesn't drive the results.
• CLV units: points for spreads/totals/props; American-odds cents for moneylines (measured from a morning-of-game baseline when no price was stated on air).
• Beat-the-close rate excludes picks whose line never moved.
• Not covered: futures, exotic/derivative markets (halves, quarters, teasers, team totals), events books didn't carry, and picks before mid-2020 (props before mid-2023) — historical odds don't exist for them.

What we track

• UnSportsmanLike with Evan, Canty and Michelle (ESPN Radio) — Play of the Day
• Sharp or Square (iHeartPodcasts and The Volume) — Full Episode
• What's Wright? with Nick Wright (iHeartPodcasts and The Volume) — Full Episode
• The Favorites Sports Betting Podcast (Playmaker and iHeartPodcasts) — Full Episode
• The Action Network Sports Betting Podcast (The Action Network and iHeartPodcasts) — Full Episode
• Get Up (ESPN) — Full Episode
• The Daily Juice - BettingPros (BettingPros and iHeartPodcasts) — Full Episode
• The Herd with Colin Cowherd (iHeartPodcasts and The Volume) — Full Episode
• Locked On Bets - Daily Sports Betting Podcast to Get The Edge (Locked On Podcast Network) — Full Episode
• You Better You Bet (Audacy and Westwood One) — Full Episode

Disclaimer

This site is unaffiliated with any show, network, or sportsbook, and exists for entertainment. It is not betting advice. Records reflect what was said on air, captured as accurately as we can.