Methodology

Quarterdeck does not try to predict game outcomes better than the betting market. That is an extremely hard problem and most services claiming to do it are guessing.

Instead it does something narrower and measurable: it compares the same bet across every sportsbook it tracks, removes each book's built-in margin, and identifies where one book is offering a materially better price than the rest of the market.

That is line shopping, and it is the most reliably profitable habit in sports betting because it requires no prediction at all — only the discipline to take the best available number.

Closing Line Value compares the price we flagged against the price the market settled on at kickoff. If we said a number was too generous and the market later moved to agree, that is evidence the read was correct — regardless of whether the bet won.

Win rate takes roughly 100+ settled plays to separate skill from luck. CLV stabilises far sooner, and it is the metric professional bettors use to evaluate themselves. A service showing you only its record on a small sample is showing you noise.

Worked example: on 29 August we flagged Jacksonville State +6.5 at +104 when the field was −105 to −120. The market closed around −105, giving a CLV of +3.13 — the price genuinely was too generous. The bet still lost 33–7. Both facts are true, and the ledger records both.

• Minimum four sportsbooks must price a market before we form a consensus.
• The required book count rises as prices lengthen — up to eight for deep underdogs, because de-vig noise grows at long prices and can masquerade as edge.
• A price gap must be at least 1.2 percentage points absolute, and between 1.5% and 12% relative. Gaps above 12% relative almost always mean our consensus is unreliable, not that we found free money.
• Most days produce few signals or none. An empty board is an honest answer, not a broken app.

• Delete or hide a losing pick. The ledger is append-only and enforced by the database.
• Change a published pick's side, price or confidence after the fact.
• Show a record without the sample size beside it.
• Use the words guaranteed, lock, or can't-lose.