Method

Two layers. The Board is where selections come from. Prime is the ticket made from them. This page sets out both, and what the numbers do and do not claim.

The Board: where a selection comes from

1. Start from the sharpest price, not from a model

Seven seasons of testing pointed the same way: no model we built beat the closing bookmaker price. When a goals model was pooled with the market, the weight the data put on the model came out negative. So there is no model here. No machine learning, no Poisson, no per-league tuning, and no AI. The starting point is the price itself.

2. Remove the margin

Every bookmaker price has a built-in margin, which is why the implied probabilities across a market add up to more than 100%. We strip that out to recover a fair probability. The result is our estimate of how likely the outcome actually is.

3. Shop the price

We compare that fair probability against roughly fourteen bookmakers and take the best price available, together with the name of the book offering it and the moment we saw it. Backtested, the same picks return around −1.6% at a typical price and +2.8% at the best price. The gap between those two numbers is where the Board’s edge sits.

Said plainly: if you take these selections at ordinary prices, at one bookmaker, you should expect to lose money slowly. The work we do is finding the price that changes that, and even then the edge is thin. We would rather write that here than imply otherwise.

4. Refuse more than we publish

A candidate has to clear several gates before it becomes a tip: the sharp book’s stake limit has to be high enough that we trust the line, the market’s margin has to be in a normal band, enough bookmakers have to be quoting it for the price to be real, and the price has to be good enough relative to fair. On a typical day most candidates fail. Every refusal is stored, with its reason, so the gates can be audited later rather than taken on trust.

Prime: the ticket

Prime is a deterministic packaging of the published Board. Nothing is picked by hand and nothing is modelled.

Prime 3 is the three highest-probability published Board tips from three different matches, in a fixed order: probability first, then the better price, then the match, then the market. Any market qualifies.
Prime 4 is exactly Prime 3 plus the next published tip from a fourth match. It is the same core ticket with one extra selection, not a second system.
If fewer than three different matches were published there is no Prime 3 that day, and it says so. If no fourth match qualifies there is no Prime 4. Blank days stay on the record.
The first leg of Prime 3 is the free leg. It is literally the first selection of the ticket, not a separate tip.
The price on a ticket is the combined price at one named bookmaker that priced every leg when the ticket was created, chosen as the highest such product. It is an observed snapshot and it may move later; it is never updated.
Every ticket is frozen before its first match kicks off, and every result stays on the record.

5. Lock it before kick-off, and publish the fingerprint

Every morning the Board is frozen and fingerprinted with a SHA-256 hash that chains to the previous day’s. Prime is frozen straight after, in its own chain, and records the fingerprint of the Board it came from. Changing an old tip or ticket would change its fingerprint and break every one after it. The database refuses edits to a locked row; only the result can be written afterwards. After a ticket settles, the frozen ticket file is released so anyone can recompute the fingerprint. This is what makes the record a record rather than a screenshot; the verify page shows how.

Historical reconstruction, and what it does not cover

Before launch, the Prime rule was replayed over seven seasons of the same market data. That historical reconstruction is context, not a live result: the dataset behind it did not contain the 1.5 goals line, while the live Board publishes many Over 1.5 selections. The historical figures therefore do not directly validate today’s exact market mix. Live Prime statistics on this site come only from live Prime tickets, starting 2 September 2026, and the calibration of every market is tracked as the record grows.

What we do not claim

That we know results in advance. Nobody does.
That the price we show will still be there when you look. Prices move constantly.
That past results predict future ones. The record includes the losing days, and it always will.