The Bookmakers' Margin at the Ebor Festival 2026

The richest flat handicap in Europe was the worst-value race of its own festival.

Across the 28 races run at York from 19 to 22 August, we measured the margin built into every starting-price book. The meeting averaged 28.1% over a fair book. The Ebor Handicap itself came in at 43.5% across its 22 runners, the highest of the week. The cheapest race to bet in was the Sky Bet Great Voltigeur Stakes (Group 2), at 11.0% with only 6 runners.

That is a spread of more than thirty percentage points inside four days at the same track, and it is not random. It tracks field size almost perfectly, which means it tracks the difference between the Group races and the handicaps.

This matters beyond bookkeeping. We published a piece midway through the meeting showing that the AI tipsters in our Silicon Tipster League made what money they made in York's Group and Listed races and gave it back in the handicaps. At the time we could describe that pattern but not explain it. The margin numbers explain it. The models were not suddenly worse at reading a handicap. They were paying roughly twice as much for the privilege of having an opinion.

Below: what a margin actually is and how we measured it, the breakdown by grade and by field size, every race ranked, and what the week would have done to somebody betting through it.

What the margin is, and how we measured it

Turn every starting price into the chance it implies. A 3/1 shot is a decimal price of 4.0, so the market is saying it wins one time in four: 0.25. Do that for every runner in a race and add the numbers up.

In a fair book the total would come to exactly one. It never does. At York last week the average race totalled 128.1%. That excess is the bookmaker's margin, and it is why a punter who bets every runner in a race is guaranteed to lose: you would be paying out more than the race can possibly return.

Two numbers describe the same thing and they are not interchangeable, so we publish both.

Overround is the amount by which that total exceeds a fair book. At 128.1%, the overround is 28.1%. This is the number most of the industry quotes and the one we use elsewhere on the site, so it is the one we lead with.

Takeout is the share of your stake the book keeps if you back the whole field: 21.9%. It is always the smaller number, because it is a proportion of the total staked rather than a proportion of a fair book.

Quoting one and calling it the other is a common way to be wrong by a third. We keep them in separate columns everywhere, and the underlying dataset has no field called simply "margin" so that nobody downstream has to guess which was meant.

The prices are settled starting prices from the published results, not the morning card. That distinction is not pedantic. The card and the settled book disagree on individual races by several percentage points, because prices move all morning and horses come out.

The Group races were cheap. The handicaps were not.

The pattern is clean. Group and Listed company sits in the high teens to low twenties. The handicaps and nurseries sit at 33.8%, and they made up sixteen of the twenty-eight races on the card.

The obvious question is whether this is about the quality of the race or the size of the field. It is the field.

Sorting the same races by runner count rather than by grade produces a tighter relationship than sorting by grade does. Group races are cheap to bet in because they are small, not because they are good. A Group 1 with eighteen runners was dearer last week than a Listed race with eleven.

There is a mechanical reason. A bookmaker sets a price on every runner and takes a slice on each one. Add more runners and there are more slices, and the ones at the bottom of the market are where the fat sits: a horse quoted at 100/1 has an implied chance of about one in a hundred, and the true chance of a horse at that price is usually longer still, and there are several of them in a twenty-two-runner handicap.

That is the whole trick of a big-field handicap. It looks like the race with the most opportunity because there are the most horses to choose from. It is the race where the market charges you the most to choose.

Every race, dearest first

Three things stand out in the full list.

The top of the table is almost entirely handicaps, and the two exceptions are the two largest-field Group races of the week. The bottom of the table is almost entirely Pattern races with single-figure or low-double-figure fields.

The Ebor Handicap sits at the very top. It is the race the festival is named after, the one with the most money bet on it, and the one where the book was hardest to beat. The two are related: a race that attracts a lot of casual money in a large field is a race where a layer can afford a wider book.

And the range is wide. The dearest race of the week charged nearly four times the overround of the cheapest. Anyone who bets a full card, or who follows a tipping service that does, is not paying an average rate. They are paying whatever the specific races they bet in happen to charge, and the difference between a shrewd card and a careless one is larger than most staking plans.

What a week of it would have done to you

A margin is a rate, not an outcome. To turn 28.1% into something you can feel, we ran the meeting again: one unit a race, win only, settled at the real starting prices, across all 28 races.

Read that table carefully, because the honest answer is more interesting than the one people expect.

Betting in line with the market costs you the takeout, and no more. If your selections mirror the market's own implied chances, your expected loss is 21.9% of what you stake. That is the mechanical floor, and it is the number the margin actually describes.

Picking at random costs you far more than that. A simulated punter choosing one runner per race at random finished the week on −51.6%, roughly twice the takeout. The reason is that random picking is not neutral. In a twenty-two-runner handicap, most of the horses are long shots, so picking uniformly means backing long shots most of the time, and long shots are where the market's margin is concentrated. Spreading your attention evenly across a field is itself a losing strategy, separately from the margin you pay.

And the distribution is not a bell curve. This is the part worth sitting with. The outcomes are skewed to the right, at 0.83, which means the typical experience is worse than the average experience. A handful of simulated punters caught a big-priced winner and finished well ahead, dragging the mean upwards, while the median punter did worse than the mean. Averages flatter a skewed distribution. The middle of the pack is below the middle of the numbers.

Of the simulated random punters, 9.8% finished the meeting in front. We should be precise about what that figure is and is not: it describes the simulation, not the public. We hold no record of what any real person staked at York last week, and we are not going to pretend otherwise by dressing a model up as a survey.

One line does finish ahead. Backing the favourite in every race returned +6.9% over the meeting. Before anyone acts on that, the same strategy over our full tested sample loses money, and four days is nowhere near enough to overturn it. That is variance, not a discovery.

Bookmakers' margin FAQ

Does a lower margin mean I am more likely to win?

No. It means you lose less slowly. Every race on the card had a margin above zero, so the market as a whole is priced for you to lose over time whichever race you pick. A cheaper book improves your terms; it does not change the direction of travel.

Why do big fields carry a bigger margin?

The bookmaker takes a slice on every runner priced, so more runners means more slices. The extra margin is concentrated at the bottom of the market, where a horse quoted at 66/1 or 100/1 has an implied chance well above its real one. A twenty-runner handicap contains several of those; a six-runner Group race contains none.

Is the Ebor Handicap unusually expensive, or is that normal for a big handicap?

It was the dearest race of this meeting at 43.5%, and the meeting itself sits towards the expensive end of the summer.

That is a seven-meeting series from one summer, not a long-run baseline, and every meeting in it is Flat. It is enough to say Ebor was dearer than most of the summer's festivals and cheaper than Galway; it is not enough to say anything about how 2026 compares with other years. Our published race-by-race results begin on 1 June 2026, so that question stays open until the series is longer.

Does taking an early price change the numbers?

Yes, and this piece does not attempt to measure that. Everything here settles at starting price. Best odds guaranteed, early prices and each-way terms all change what a real punter pays, sometimes substantially. They are separate questions and each deserves its own measurement rather than a hand-wave here.

Where do these numbers come from?

Every figure resolves live from a frozen dataset record for the meeting, built from the settled published results. Nothing on this page is typed in by hand, which is deliberate: figures that are typed once and left alone are the ones that quietly go stale. The methodology, the schema and the conventions are written up in how we measure a festival.

Why publish this at all?

Because the margin is the one number in betting that is knowable in advance and almost never quoted. Prices are published, results are published, and the gap between them is arithmetic. A punter who knows that a big-field handicap costs roughly twice what a small Group race costs is better equipped than one who does not.

Bet responsibly. Nothing on this page is a tip or betting advice. If you bet, stake only what you can afford to lose, and if it stops being fun, stop. Help is at BeGambleAware.org. 18+.