The Math Behind the Magic: How vibeodds Calculates Fair Odds

One of the most common questions we get is: "How do you know what the 'Fair Odds' are?"

It's a fair question. If our fair odds are wrong, the Expected Value (EV) is wrong, and everything built on it falls over. So here is the methodology: not the proprietary code, but the actual reasoning, including where it's strong and where it isn't.

TL;DR: We don't predict horses; we measure the market. Fair odds come from aggregating every bookmaker and exchange price in real time, stripping out the bookmaker margin (overround), and weighting the sharpest sources — above all the exchange, where real opposing money sets the price. A bookmaker offering odds above that margin-free consensus is out of line with the smartest available estimate, and that gap is the EV% you see.

The wisdom of the crowd

We don't try to predict the future. We don't look at the horse's form, the jockey's diet, or the ground conditions.

Instead, we look at the market.

The betting market is a highly efficient information processing machine. Millions of pounds are traded, and thousands of smart people (and algorithms) are constantly adjusting prices based on every piece of available information. Any single input (form, going, gallop reports) is already inside the prices, weighted by people risking money on their reading of it.

1. Aggregating the market

We scrape odds from the major bookmakers and exchanges in real time. This matters for two reasons: a consensus built from many books is harder for any one compiler's error to distort, and the spread between books is itself the raw material. Value hunting is precisely the search for the book that disagrees with everyone else.

2. Removing the "overround"

Bookmakers build a profit margin into their odds (the "vig" or "overround"). If you sum the implied probabilities of all horses in a race, it might add up to 115% instead of 100%. That extra 15% is the bookie's edge; our overround guide covers it fully.

A concrete miniature. A three-horse race priced 2.00 / 3.00 / 5.00 has implied probabilities of 50% + 33.3% + 20% = 103.3%. Strip the 3.3% margin proportionally and the fair probabilities are roughly 48.4%, 32.3% and 19.4%: fair odds of about 2.07, 3.10 and 5.17. Every price in that book was slightly worse than fair; the margin removal shows exactly how much.

One subtlety separates a serious model from a naive one: the margin isn't spread evenly across the field. Bookmakers systematically shade outsiders harder than favourites (the favourite-longshot bias): a 50/1 poke carries proportionally far more margin than the 2/1 market leader. Removing the overround with allowance for this, rather than pro-rata, is one of the differences between fair odds that work and fair odds that flatter longshots.

3. Weighting the sources

Not all prices are equal.

The exchange is the anchor. Exchange prices are set by opposing punters matching real money, with no built-in margin, the closest thing racing has to a true probability, especially in the liquid final minutes.
Sharp books react fast and take serious money; their moves carry information.

Soft books are slower to react and more prone to being out of line — which makes them the

input to a fair-odds model and the

Our model leans on the sharp end. If the sharpest money in the world says a horse is a 4.00 shot, and a soft book still has 6.00 up, that 6.00 isn't a difference of opinion worth averaging in. It's the opportunity.

The result: a source of truth

By blending these inputs, we create a fair odds line representing the market's best margin-free estimate of each horse's chance.

We then compare this line against every available bookmaker price:

Fair odds: 4.00 (25% chance)
Bookie odds: 5.00 (20% implied chance)
EV: 5.00 ÷ 4.00 = 125%, and that percentage is exactly what the EV column shows.

Why this works — and what it can't do

This approach beats solo form study for one structural reason: it borrows the collective intelligence of the entire market. No individual analyses information faster or better than everyone with money at stake combined. We simply find the spots where individual bookmakers have drifted away from the consensus — fallen asleep at the wheel for a horse, a race, a morning.

Honesty about the limits, because a model you can't trust the edges of is worthless:

Thin markets are noisy markets. A Monday midwinter maiden an hour before the off has little exchange liquidity, and the "consensus" is a handful of small bets. Fair odds are strongest where money is deepest: big races, near the off. EV flags in thin markets deserve more scepticism than the same number in a deep one.
The model is only as current as the prices. When news breaks — a going change, a non-runner — every source is repricing at once, and for a few minutes the consensus itself is in motion. Some of the largest EV figures appear in exactly these windows; some are real (a slow book), some are artefacts (a stale snapshot). Fresh quotes decide which.
A persistently "wrong" book might know something. Occasionally the outlier is ahead of the market, not behind it. This is rare — the consensus wins the argument the vast majority of the time, which is the whole basis of the method — but it's why value betting is played as a portfolio across many bets, never as certainty about one.

The proof, in the end, is empirical: if the fair odds are honest, bets flagged at 115% EV should settle, over hundreds of results, near +15% ROI. Tracking realized return against predicted EV is the standing calibration test, and it's the comparison the Results view exists to make visible.

See the model in action on the Live Odds page and in the Value view, where fair odds, best price and EV% sit side by side. Then follow those prices through to the Results view to see how often the market — and your decisions — were right.

Frequently asked questions

Why use the exchange as the anchor rather than an average of bookmakers?
Because exchange prices carry no margin and are set by money on both sides of every price. An average of bookmaker prices is an average of margins and marketing decisions; the exchange is the cleanest probability signal available.

If the market is so efficient, how does value exist at all?
The consensus is efficient; individual books are not. Dozens of firms price thousands of runners daily under commercial pressures (promotions, liability limits, slow trading desks). The consensus is the ruler; the value is each book's deviation from it.

Does the model know about going, draw, or trainer form?
Only through the prices. Everything public is already in the market — that's the premise. What the model adds isn't racing knowledge; it's the margin-stripping and cross-book comparison no eyeball can do at scale.

Why do EV figures sometimes vanish within minutes?
Because the out-of-line book repriced, or the consensus moved to meet it. Value is a perishable disagreement, not a property of the horse — the drift guide covers what those moves mean.

What EV% should I actually trust?
Moderate figures in liquid markets are the reliable core. Enormous EVs in thin markets or chaotic minutes deserve a second look before money — see interpreting EV for the full sweet-spot argument.