About
Arithmetic on public prices
The idea
Every bookmaker's odds contain a margin. Add up the implied probabilities of a market's outcomes and you get more than 100% — the excess is the bookmaker's cut, and it is why betting at random loses money slowly and reliably.
Pinnacletakes a smaller cut than almost anyone, accepts far larger stakes, and does not restrict customers who win. Those three facts force its line to be accurate: a book that welcomes sharp money and prices badly gets taken apart. That makes Pinnacle's line the best public estimate of true probability available — once the margin is removed from it.
Removing that margin is the whole method. What is left is a fair price. Any bookmaker offering more than the fair price is, on average, paying you too much — that is a value bet. And when two books disagree enough that every outcome can be backed at a profit, that is an arbitrage. Both are division. Neither is a prediction.
How it is built
A Go engine crawls every listed fixture across the covered books continuously, matches the same fixture across them by name and kickoff, prices each market, and writes the results to a database the web app reads. Scans are not on a fixed schedule so much as a continuous loop — a full pass over the reference book's listing, then the next one.
The parts worth knowing about, because they are where the accuracy actually comes from:
- Fixture matching. Books spell teams and leagues differently. Matching is fuzzy name similarity plus kickoff time, backed by learned alias registries so a pairing confirmed once stays confirmed. A mismatched fixture is the most expensive error possible, so low-confidence pairs are excluded from arbitrage entirely.
- Settlement rules. Books settle the same market differently — Pinnacle gives a tennis moneyline action once a set is complete, while others void on retirement. Legs only combine within a matching rule class, and a book whose rules have not been checked is excluded rather than assumed.
- Staleness. Prices carry their age. A book that stops updating is dropped rather than quoted, because an unmaintained price looks exactly like an opportunity to anything ranking by edge.
- Closing line value. Every bet taken is graded against the closing price. That is the standard evidence that an edge was real rather than lucky, and it is the number that matters long before profit does.
What we refuse to do
This market is full of tools that advertise a number. The useful thing to publish is what the method cannot do, because everything else is checkable and this is the part a reader would otherwise learn the hard way.
- No tips and no predictions. There is no model. A +6% bet loses about as often as its odds say it should.
- No claimed returns. The figures on the marketing pages are worked examples from the prices printed beside them, labelled as such. Your record is your own.
- No hidden coverage. Four bookmakers and one exchange, across soccer, tennis and basketball. That is stated on the pricing page next to what the paid tools cover, including where they cover more.
- No pretending the ceiling does not exist. Soft bookmakers restrict accounts that consistently beat the price. Staking sensibly lengthens the runway; nothing removes the limit.
This is a research tool, not a living
Expected value is an edge measured across hundreds of bets. Variance inside that is large enough to look exactly like being wrong for months at a time. Anyone telling you otherwise is selling something.
Contact
Questions, corrections and bug reports are all welcome — particularly corrections. If something on this site is wrong, tell us and it gets fixed.
Email goldedge.net@gmail.com, or see the contact page.
For research, not betting advice. Positive expected value is an edge across many bets, never a prediction about one. Bet only what you can afford to lose. 18+.