How it works

What you get

A price on every match in the Premier League, Championship, League One, League Two, La Liga, Serie A, the Bundesliga and Ligue 1. Not a shortlist. Not the three we fancy this weekend. All of them, every result, published before kickoff and never touched afterwards.

What we publish is a fair price: what the bet is genuinely worth, with nothing taken out of it. The price you get from a bookmaker has their cut built in, which is how they pay for the shirts and the shop fronts. That cut is real money and you almost never get to see it.

Because nothing here is picked, whatever you were going to bet on is already on the board. You are not waiting for someone to tip it.

What to do with it

Use our price as a line, and hold whatever you are being offered up against it.

Say we make Everton to win worth 2.30. If your bookmaker is offering 2.10, you can see what they are charging you for that bet. A tenner returns twenty-one quid instead of twenty-three.

If someone is offering 2.50, the price is longer than we think the bet is worth, and that is one worth a look. A tenner returns twenty-five.

That second one is the useful half, and it does not need us to be cleverer than the bookmaker. Prices on the same result differ from one shop to the next, all the time, for no better reason than who has taken what. Knowing roughly what a bet is actually worth is what tells you which of them is the one to take.

Worth a look is doing honest work in that sentence. It is only a good price if our number is right, and there are no certainties in any of this. What we can do is show our working on that, which is the record page, and let you decide how much weight it carries.

We show one bookmaker's price next to ours, exactly as they posted it, cut and all. Where you actually bet is your business.

Where the number comes from

We work out how many goals each side is likely to score, then build one picture of the match: every scoreline it could plausibly finish, and how likely each one is.

Everything else comes off that same picture. Both teams to score, over or under two and a half goals, the match finishing 2-1: those are not three different opinions that might end up contradicting each other. They are three questions asked of the same thing. Plenty of tools get that wrong, and you would never know from looking.

Two things go into the goals.

The first is how good the sides are. Every club carries an attacking and a defensive rating, built from years of results, with what happened last month counting for more than what happened two seasons ago.

The second is who is playing. We build the price around the eleven we expect to start, and each carrying a rating from the sides they have played in. A team missing three of its first-choice starters does not get priced as though they were all fit.

Then the part that matters most and that almost nobody else does. Left alone, a model like this is quietly overconfident about favourites and too shy about draws. So we go back over everything it has predicted before, compare it against what actually happened, and correct it. That correction is worked out separately for each league and each market, because a Championship goal line goes wrong in a different way from a La Liga home win. If you want the name for it, it is isotonic regression, and if you do not, it is just marking our own homework and fixing what we got wrong.

One honest exception. The scoreline grid, the one showing 1-0 and 2-1 and the rest, does not get that correction. It is there to show you the shape of a match rather than to tell you that 2-1 has been checked and stands up.

What it will not do

It will not tell you what to back. There is no bet of the day here and there never will be.

It will not tell you a bet is value. On the big leagues our prices land pretty much where the bookmakers land once you strip their cut out, which means the market is doing its job rather than us finding something. There are divisions we have not looked at closely enough to say either way. If that changes we will tell you, and we will show you why.

It will not get excited about a run of form. Recent results already count for more than old ones, and who has been playing shapes who we expect to start. A team winning four on the bounce that has not actually got better does not get a better price for it.

It does not look at what the bookmakers are doing. Feeding their prices in would make our numbers look sharper and would wreck the only thing this page is claiming.

Marking our own homework, in public

The record page asks one question of everything we have ever published. When we said something was thirty percent likely, did it come in about three times in ten?

You can see the answer band by band, including the bands where we are off, because a page that only showed the flattering ones would not be worth reading.

We do not put a big accuracy number at the top, and that is not modesty. We price every result of every match, and exactly one of them always happens, so on a match result our strike rate is one in three before a ball is kicked. That is a fact about football having three outcomes, not about us being any good. A tipster's strike rate means something because a tipster picks. We do not pick.

And nothing moves after the fact. Every price is stamped with the moment it went up and cannot be edited afterwards, not by us either. What you see on a match that has already been played is exactly what stood before it kicked off.

The bit at the bottom

This is a tool for working out what a bet is worth. It is not tipping, it is not advice, and it is not a promise about any single match. Bookmakers take a cut and most people who bet lose money. PunditModel is not a bookmaker and takes no stakes. 18+.