Where the numbers come from

The data

Prices come from The Odds API, restricted to the UK region, which returns 21 UK-licensed bookmakers and exchanges. We cover 49 football competitions and two markets: match result and over/under 2.5 goals.

Only two markets, deliberately. Those are the two every UK book quotes on every fixture, so the comparison is like for like across all of them. A market only six books quote would produce a “best price” that quietly means something different.

Live pages are cached for ten minutes. Every page states the time its prices were read. Close to kick-off, prices move faster than that, so treat a price here as a strong indication of where the market is and confirm it with the bookmaker before betting.

Implied probability

Implied probability is 1 ÷ decimal odds. A price of 4.00 implies 25%. Summed across every outcome in a market this comes to more than 100%, and the excess is the bookmaker’s margin. We show that sum as “book %” on every fixture page.

Removing the margin

To get from published prices to a probability that sums to 100%, we divide each outcome’s implied probability by the total. This is proportional de-vigging, and we use it for one reason: you can reproduce it on a phone calculator.

It has a known flaw. Because bookmakers load more margin onto longshots than favourites, proportional de-vigging overstates the true chance of outsiders. Methods like Shin or the power method handle that better, and we do not use them, because a number nobody can check is worse than a number with a stated limitation.

Best price and best versus average

The best price is simply the highest price any covered bookmaker is publishing on that outcome. Best versus average compares it to the mean of every book quoting the same outcome. An outcome is only shown at all if at least three books quote it.

Exchange prices are included and marked. Betfair, Smarkets and Matchbook charge commission on winnings, so their headline price overstates what you keep by roughly the commission rate.

The kick-off price archive

We capture prices repeatedly as kick-off approaches and store every capture. The last capture inside three hours of kick-off is treated as the closing price. If we have no capture inside that window, we say so rather than substituting an earlier one.

The archive currently holds 107 captures across 66 fixtures, starting 2026-09-05.

Closing line value

For a logged bet, we compare the price taken against the same market’s price at kick-off, after removing the margin from both. Working on margin-free probabilities rather than raw prices matters, because a raw price can look better purely because that book runs a thinner margin.

Consistently taking prices above the closing line is the clearest available evidence that a bettor is finding something the market has not priced. It is evidence, not proof, and it is worthless on small samples. Our tracker keeps showing you the sample size for that reason.

What we tested, and what failed

This product is shaped by research that mostly produced negative results, and it is more useful to publish those than to hide them.

• Around 70 predictive hypotheses were tested across 39,621 fixtures, 18 leagues and 13 seasons of results and closing prices, using chronological train and test splits with correction for multiple comparisons. No strategy survived out of sample.
• A price-dispersion flag, best price at least 5% above the market average, measured -0.97% against the closing line. It is a longshot filter, not a signal, and it was withdrawn from publication.
• A backtest of accumulators built from independently priced legs returned -32.9%, dramatically worse than the singles they were built from.
• The favourite-longshot bias is real in this data: favourites win between one and four and a half percentage points more often than the de-vigged price implies. It is smaller than the roughly 5.5% margin, so it cannot be turned into money.
• One finding did survive pre-registered testing: prices at softer books that had drifted out of line with the market carried +4.89% closing line value with a confidence interval of +2.20% to +7.53% across 1,089 samples. That is line shopping. It is why Touch Line Bets is a price comparison and record product rather than a prediction product.

What Touch Line Bets cannot tell you

It cannot tell you who will win. It does not try. There is no model probability, no rating, no confidence score and no selection anywhere on this site, and there never will be.

It cannot promise that taking better prices makes you profitable. Better prices make a losing strategy lose more slowly and a winning one win more. They do not turn one into the other.

And it cannot protect you from the structural problem in UK betting: the Gambling Commission found that 4.31% of active betting accounts carry commercial restrictions, roughly 650,000 of 14.9 million, and that 46.78% of restricted accounts are in profit. Bookmakers restrict customers who win. No tool changes that.

See the prices →