The value feed — every edge priced against the exchange

Each selection is a complete object: the bookmaker and its offered odds, the market and the selection, and the fair price it is measured against — always the exchange, stated as the fair_source, with the exchange named and its commission applied. From those two prices the feed derives the implied and true probabilities, the edge, the EV percentage and a Kelly fraction, so your product filters and stakes rather than computing the maths itself. The docs carry the field-by-field response and the coverage dashboard shows the books and freshness behind it.

Fair means fair against something real. The value here isn't a house model's opinion — it's the gap between a bookmaker's price and the current exchange price for the same selection, de-vigged and net of commission. That is why the feed is exchange-based rather than model-based: the benchmark is a live market other people are betting into, not an estimate. The exchange-lay page covers where that exchange side is priced.

Most value data on the market is raw prices with a modelling problem attached: normalise the books, build a fair-odds benchmark, resolve the same selection across sources, compute the edge, keep it alive as prices move. The value feed does that computation over OddsRelay's own board and hands you the result — the selection, its edge, its EV% and its Kelly fraction — so the endpoint is the input to your staking and alerting logic, not its first month of engineering. The expected value API spoke walks the developer view.

Coverage decides how much value there is to find. The board spans 60+ UK bookmakers with bet365 included as standard, and the fair side is priced from Betfair, Smarkets and Matchbook — so a soft price at a book a shallow feed skips is still an edge your product can surface. Filter by minEv to set your value floor, cap position size with maxKelly, and narrow to the markets or bookmakers you act on.

A +EV feed needs a fair price to measure against, and here that price is the exchange — so no exchange means no fair price, and no honest +EV. It is live across UK and Ireland today, the markets with the exchange liquidity the benchmark depends on. Where a market has no in-region exchange, we don't substitute a house model and call it value: the regions view states plainly which markets carry a +EV surface and which don't. We never market a +EV feed in a market that has no exchange to price against.

That gate is the same one the exchange-lay page and the regions surface enforce — a deliberate limit that keeps every edge you see anchored to a real, live benchmark rather than an opinion.

A positive EV is a modelled expectation over many bets, not a guaranteed win on any one — say it plainly and it stays trustworthy. The edge is the gap between a book's price and the exchange fair price at the snapshot instant; variance is real, prices move, and the fair benchmark is itself a live market that can shift. The Kelly fraction is a sizing guide against that expectation, not a certainty. A good feed gives your product more edges, priced against a real benchmark, fresh — it doesn't remove variance, and we won't imply it does.

Two more limits up front. The feed is pre-match on a ~3s polling cycle — right for pre-match value, not in-play. And it computes +EV over the current board today; full cross-book +EV across all 60+ books at once, streaming delivery, historical archives and dropping-odds or closing-line-value views are on the roadmap, described as coming rather than shipped. If you'd rather test than take our word, start a free trial and point your staking logic at the live endpoint this week.