How it works
What the four methods actually say
Stripping a book's margin is a modelling choice, not arithmetic. On the -145 / +125 market the four standard methods land here — and two of them, additive and shin, land on the same number, which is the sort of thing a table of four decimals hides.
The consensus is 57.34% and the spread is 0.39 points. That sounds small until you put it against the width of an edge worth having:
What rounding a stake costs
The exact arbitrage stakes are 493.98 and 506.02. Nobody who is not running a tool bets 493.98, and stake precision is one of the cheapest signals a risk desk has. The round pair sits underneath:
Rounding costs 2.35 of the 37.35 locked if both bets are accepted and stand — the notch below:
How old the price already is
A quote that looks 10 seconds old is really 15.4, because the feed took 5.4 seconds to reach you and those seconds were invisible. The hatched region is the part nobody showed you:
Fill probability follows the real age, not the shown one: 89% becomes 84%, and a 4.0% screen edge is worth 3.37% rather than 3.58%.
Welcome offers, after every cost
A bonus bet does not return its stake, so it is worth about half its face value if bet naively. Converting means hedging — and every guide says “convert on a longshot” without mentioning what that costs at the second book. On a 1,000 bonus:
| Free leg | Hedge at | Hedge stake | Locked if both stand | Rate |
|---|---|---|---|---|
| +100 | -110 | 524 | 476 | 47.6% |
| +200 | -230 | 1,394 | 606 | 60.6% |
| +400 | -450 | 3,273 | 727 | 72.7% |
| +1200 | -1400 | 11,200 | 800 | 80.0% |
The hedge column is the constraint the advice always omits.
Middles, counted rather than assumed
Results are not smoothly distributed. Football margins pile up on 3 and 7 because of how the sport scores, so a normal curve gets the shape wrong in exactly the place the whole bet lives. Same prices, one point apart:
| Window | Hits | Break-even | Expected value |
|---|---|---|---|
| 2.5 / 3.5 | 15.79% | 4.76% | +10.53% |
| 4.5 / 5.5 | 4.21% | 4.76% | -0.53% |
Computed from a 285-game sample whose shape is synthetic, shape stated — an illustration of the mechanism, not a claim about any league. No published margin table we could verify covers every margin with its sample size, so Bookbreaker counts the games you record rather than shipping a distribution, and refuses to give a counted answer below 200 games.
Which books you can use
27 venues, filtered to the ones you can actually hold: an arbitrage between two books you cannot both open is noise with a number attached. 12 in New Jersey, 2 in Florida. Table read 2026-08-17, and the site shows its own age because a jurisdiction table without a date asserts “true now” forever.
A dated starting point for your own check, not advice. The book's own site decides whether it will accept you.
What stays out of the product
It does not link sportsbook accounts. Trackers that sync automatically do it by holding your sportsbook credentials; Bookbreaker imports the CSV your book already exports and never asks for a login.
Nor does it do multi-accounting, identity or KYC workarounds, or device and location spoofing — see account longevity for where that line is drawn and why it is drawn in code rather than in a policy document.