About

Matchday Price is a public record of football data. The prices bookmakers are offering, the results when the matches finish, and the numbers counted from them. Free to read, no account, and every figure carries the date it was taken and the sample it came from.

That record is the thing itself. It doesn't depend on any theory being right, and it's worth reading whether or not anything else here works out.

What sits on top of it are experiments. Two so far. Both are open tests rather than products, and neither is a claim that there's money to be made.

The record

The board carries every fixture in the next three weeks across eight competitions, with the best price held for each result and for over and under 2.5 goals, the bookmaker offering it, and the probability the rest of the market implies. Where no price is held, the fixture is listed anyway and says so.

Open any match and you also get both clubs' records at the venue they're playing at: goals scored and conceded, corners won and conceded, cards, and how often their matches finished with three goals or more. A side's home form and its away form are frequently two different teams, and a single figure hides that.

The league numbers are counted from every finished match held, across four seasons. No model is involved in any of them. Every rate carries the number of matches behind it, and where a sample is too small to mean anything it says so rather than printing a percentage.

First avenue: Fair Price

Ask a bookmaker for two selections on one betslip and you're usually quoted the two prices multiplied together. That only works if the two events are independent. Football outcomes frequently aren't, because they're driven by the same thing: how many goals each side scores.

Fair Price works out what those combinations should cost using a model that accounts for the relationship, and posts the result before kickoff so a record exists that can be judged later.

What's being tested is narrow, and worth stating exactly. Given the same market-derived probability for each individual selection, does a correlation-aware model predict the combined outcome better than multiplying those same probabilities? Both methods start from identical inputs, so correlation is the only thing that differs between them.

It is not a test of whether bookmakers get combinations wrong, however tempting that framing is. Testing that would need the price a bookmaker actually offers for the combination itself, and no such price is held here. The odds feed supplies each market separately and nothing about same-match combinations, so that comparison can't be made and isn't claimed.

Nothing has been posted yet, and there's nowhere near enough settled data to say whether the idea works. It may turn out that it doesn't.

Second avenue: Second Look

The board holds prices and counted history. It holds nothing about whether four defenders are injured, who's suspended, what the weather is doing, or that a manager was sacked on Tuesday. A fixture where a side is missing half its first team looks exactly like a fixture where everyone is fit.

Second Look is the plan to put those things next to the numbers. Not to produce a selection, and not to hunt for an edge, but so that someone reading a fixture can see what the numbers don't know before they decide what to make of them.

None of that data is held yet and nothing has been built. It's named here because it's the next avenue, and because saying what's intended belongs to the same discipline as dating what's published. When it exists, it'll arrive with its sources and its gaps stated like everything else.

Why neither of these is an edge claim

Both arms could fail. Fair Price could show that multiplying was fine all along. Second Look could show that knowing about the injuries changes nothing you could act on. Those are real possible outcomes, and if either happens it gets reported as plainly as anything else here.

That's the reason the record comes first and the experiments come second. A counted, dated, checkable record of what the market said and what actually happened is useful on its own terms. Everything else is us finding out.

Why the dates matter

Anyone can publish a winning record after the results are in. The only way to show a claim was made beforehand is to have someone independent attest to it at the time.

So every time this site is published, a list is written of every page along with a fingerprint that changes if a single character of it changes. That list is then lodged with two services that keep their own dated copies: a public timestamping service, and the Internet Archive.

Neither is ours to go back and edit. Which means you can check any claim here against what the site actually said on the day, yourself, without trusting us. The publication record explains how to do that.

The same rule applies inside the site. Nothing appears on the Fair Price combinations page unless it was priced before kickoff, and anything priced close to kickoff is shown with a marker attached rather than quietly dropped.

Who this is for

People who bet, mostly. Nothing here tells you what to back and nothing is sold, but the prices, the form splits and the base rates are the things you'd want to look at before deciding, and they're gathered in one place with their dates and samples attached.

If you're here for tips, there aren't any, and there won't be.

Responsible gambling

This site is a record of data and predictions, not betting advice. If you do bet, bet responsibly. 18+ only. Help is available at GambleAware.