Book Accuracy
What it is
Book Accuracy is the calibration dashboard. We graded 12M+ historical lines across 74 sportsbooks, then measured each book's calibration error: how far their implied probabilities deviate from actual outcomes. The result is a per-sport, per-market accuracy matrix. Sharp books cluster at the top. Soft books at the bottom. The bottom is where your edges live.
When to use it
- Before placing a bet, to confirm the offering book is actually soft on this market.
- When prioritizing which books to open accounts at (start with the softest).
- When investigating an edge that looks too good (a Polymarket-only edge with no Pinnacle anchor is suspect).
How to read it
| Field | Meaning |
|---|---|
| Book | Sportsbook name. |
| Sport / Market | Calibration is sport- and market-specific. Book may be sharp on NFL spreads but soft on NBA props. |
| Sample Size | How many graded lines we have. Bigger = more reliable. |
| Win Rate vs Implied | If a book quotes 55% favorites that win 53% of the time, they are 2 points soft. Higher gap = softer. |
| Calibration Error | Mean absolute deviation between implied and realized probability across all bins. |
| Tier | S (sharpest), A, B, C, D (softest). Drives our v4 weighting. |
| Vig | Average hold percent. Sharp books have low hold. Some soft books also have low hold; do not confuse the two. |
Worked example
Example reading
Book: Pinnacle | NFL spreads | n=180,000 | win rate gap +0.2% | tier S | hold 2.0% Book: Circa | NFL spreads | n= 22,000 | win rate gap +0.4% | tier S | hold 4.5% Book: FanDuel | NFL spreads | n=420,000 | win rate gap -1.8% | tier C | hold 4.5% Book: Fliff | NFL spreads | n= 35,000 | win rate gap -3.4% | tier D | hold 6.0%
Pinnacle and Circa are nearly perfectly calibrated (tier S). FanDuel is 1.8 points off in your favor when you take the dog. Fliff is 3.4 points off. Edges on Fliff are real and big. Edges on Pinnacle are usually arbs, not value.
Common mistakes
- Treating accuracy as global. A book can be sharp on one sport and soft on another. Always read the per-market row.
- Ignoring sample size. 1,000 lines is not enough. Trust S/A tier rankings only when n is large.
- Assuming low vig means sharp. Some retail books quote tight to look credible. Sharp = calibrated, not just tight.
- Skipping the check before placing. An edge at a tier-S book is suspect. An edge at a tier-D book is the norm.
- Forgetting that books recalibrate. Soft books get sharper over time as they hire risk teams. Recheck quarterly.