What is positive expected value betting?

Almost every piece of betting advice you have ever read is about picking winners. Form guides, injury news, a tip from someone who follows the team.

Expected value betting throws that out. It does not ask who will win. It asks a narrower and far more answerable question: is this price wrong?

The difference sounds academic. It is really the difference between a hobby that slowly costs money and a process that can be measured.

Start with a coin

A fair coin lands heads half the time. Nobody disputes this, which is exactly why it is useful. With the probability settled, the only thing left to argue about is the price.

A bet on heads at odds of 2.00 is perfectly fair: stake $10, win $10 half the time, lose $10 half the time, break even forever.

Now change nothing about the coin and change only the number. At 1.90 you lose slowly. At 2.10 you win slowly. The coin has no idea which you took.

Below are two bets on the same flips. Not two simulations, but one sequence of coin flips priced two different ways. Flip it once. Then flip it a thousand times.

What that shows, and what it does not

Three things are worth taking from it.

One flip tells you nothing. At ten flips the cheap price is often ahead. This is the single most common reason people abandon a good process and keep a bad one. They judge it on a sample far too small to show anything.

The gap is not luck, and it does not average out. Both lines got identical outcomes. Everything separating them is the price, applied to every single stake, compounding.

Being right is not the same as being paid. You can pick the winner and still lose money over time if you keep taking 1.90 about it.

That last point is the whole discipline. It is also why we do not lead with a win rate. A win rate can be made to look like almost anything by choosing which odds to bet at.

Where the price comes from

A coin is easy because everyone agrees it is 50/50. A rugby match is not.

So the practical question becomes: where do you get a probability you can trust more than the bookmaker's?

The answer is not a model in a spreadsheet. It is other bookmakers, and specifically the most accurate ones in the world. A handful of them (Pinnacle is the one most people have heard of) take enormous amounts of money, welcome customers who win instead of shutting them down, and correct a price within seconds of it being wrong. Their prices include a cut of their own, but take that cut out and what is left is the best available estimate of what a result is really worth.

Compare that cleaned-up price to what a local bookmaker is offering, and the gap is the edge. That is the entire method. It is arithmetic, not prophecy.

It is also why the size of the local cut matters so much. We measured TAB's. It runs at a median of 7.2% across 15,131 markets, and the full breakdown is here. That is the hole any edge has to climb out of before you make a dollar.

The problem with waiting

Here is the uncomfortable part of everything above. If a real edge is a couple of percent, and results swing wildly for hundreds of bets, then you cannot tell whether a betting process works by looking at whether it is winning. Not for a very long time. Hundreds of bets is months. By then you have either quit or gone broke on something that was fine, or happily kept going on something that was not.

So how does anyone know sooner?

A faster test: did the price move toward you?

There is a much quicker check, and it is the one people who bet for a living actually use.

Just before an event starts, its final price has absorbed everything: every injury, every team sheet, all the money that was ever going to be bet. That final price is the sharpest estimate anyone will ever produce about that event.

So the question is simply: was the price you took better than that final price?

If you backed something at 2.40 and it finished at 2.10, everyone else moved toward your opinion after you had already bet. You bought at a better number than the market eventually settled on. Do that consistently and you are finding prices before they get corrected, whether or not that particular bet won.

This resolves far faster than profit does, because it uses every bet instead of waiting for winnings to separate themselves from luck. It is the difference between "we feel like this is working" and a number you can audit.

How fast is "faster"? It depends on how many bets, not on the calendar. A market we only send a few picks a month in still takes months to say anything. The point is that the same number of bets tells you far more when you score them on price than when you score them on the result.

It also cannot be faked by picking favourable examples. A losing bet whose price moved your way still counts as a good bet. A winner you took at a worse number than the market settled on still counts as a bad one. That is uncomfortable, which is rather the point.

What we do with this

Our scanners price TAB against the most accurate bookmakers in the world every fifteen minutes, and flag the prices that are clearly wrong. That is the coin-flip widget above, run continuously across thousands of real markets.

Then every one of those picks is scored against that final price, in public. Not the win rate, but the price. Everything we have sent since, won or lost, sits on the track record with its final price beside it.

"Since" is doing real work in that sentence, so here is what it means. The record shows every pick from the markets we still publish. It leaves out markets we stopped sending, which were a stretch of club friendlies and low-tier tennis we no longer touch, plus some early bets entered by hand. A chart drawn over those would answer "how did every idea we ever had go?", and the only question worth answering is "how does what you actually get sent go?". Ninety-one settled picks sit outside the record for that reason.

You do not have to believe the method. You can check it.

Before you take any of this too seriously

Expected value is not a way to win bets. It is a way to make sure the bets you take are priced in your favour, and then to survive the enormous swings between here and the long run.

Losing runs are normal and they are longer than you think. The edge only shows across hundreds of bets. How much you stake matters more than which pick you take. And none of this makes betting a sensible way to make money. It makes it a measurable one. Never stake more than you can afford to lose.

Happy betting, Seb

Why I built this, and the broke-student version of the story, is on the why page.