You’ve probably read a glowing system writeup that boasted a 32% strike rate and a small profit at SP. Pretty graph. Convincing tables. Then a year later you ran the same filter… and it bled.
The reason is almost always the same. Strike rate and ROI tell you how often you won and how much you made. Neither tells you whether you were paid the right price. The number that does is the A/E ratio: the actual winners divided by the winners the odds said to expect. It’s the first figure I put next to any betting angle before I trust it.
The short version
- A/E of 1.00 means the prices were fair. Above 1.00, the horses won more often than the odds implied. Below 1.00, they won less often.
- Work it out against Betfair SP, because bookmaker prices carry a bigger margin.
- Small samples fool everyone. The fewer bets behind an A/E, the less it can tell you.
What is an A/E ratio in horse racing?
A/E stands for Actual divided by Expected. It compares how often a group of horses actually won with how often their odds said they should win.
| A/E | What it means |
|---|---|
| 1.00 | The prices were right |
| Above 1.00 | The horses won more often than the odds implied, so there was value |
| Below 1.00 | The horses won less often than the odds implied, so the prices were too short |
How do you calculate an A/E ratio?
The maths is short:
- For every horse in your sample, turn the decimal odds into an implied probability by dividing 1 by the price.
- Add those probabilities up across the whole sample. That total is your Expected number of winners.
- Count the actual winners in the same sample.
- Divide actual by expected.
A worked example makes it concrete. Say you’ve backed 200 horses, all priced at 4.00 on Betfair. Each has an implied chance of 1/4.00, which is 0.25, or 25%. Across 200 bets, the market said you should win 200 x 0.25 = 50 races.
If you won 56, your A/E is 56 / 50 = 1.12. The market underrated your selections by 12%. If you won only 44, it’s 44 / 50 = 0.88, and the market overrated them by 12%.
One detail matters. Work it out against Betfair SP, not bookmaker SP. Bookmaker prices carry a bigger margin, which makes the implied probabilities add up to more than 100%. That inflates the Expected winners and pushes A/E down artificially. A/E uses the raw prices. Betfair’s 2% commission on net winning markets only matters later, when you turn it all into profit.
Why do strike rate and ROI mislead where A/E doesn’t?
Strike rate tells you how often. ROI tells you how much. Neither tells you whether the price paid you fairly for the risk. A/E does.
Imagine two systems:
| System A | System B | |
|---|---|---|
| Bets | Odds-on favourites at an average price of 1.80 | 14/1 outsiders |
| Wins | 58.5% | 7.0% |
| Expected wins from the prices | 55.6% | 6.7% |
| A/E | 1.05 | 1.05 |
Their strike rates look incomparable. The A/E shows they carry the same edge, one expressed through short prices and the other through long ones. Without A/E you’d have no honest way to compare them.
The same idea exposes systems that look like winners but aren’t. A backtest can show a small profit while running an A/E of 0.92, because luck has carried it. Run that filter forward for another year and the thin prices tend to win in the end.
What do 13 years of UK racing show about A/E?
I looked at every UK race from July 2013 to March 2026. PR and VDW are my own ratings, and Race Predictor turns the evidence on each runner into a win probability. Each row below is the top-ranked horse on that measure. Here are the headlines:
| What I tested | Win rate | A/E |
|---|---|---|
| Betfair favourite (162,036 races) | 34.4% | 1.00 |
| Top Official Rating in handicaps (132,481 runners) | 14.3% | 0.99 |
| PR Rank 1 (159,771 runners) | 32.3% | 0.99 |
| VDW Rank 1 | 25.6% | 0.99 |
| PR Rank 1 and the top Official Rating are the same horse, handicaps (25,429 runners) | 29.0% | 0.99 |
| Top-ranked by Race Predictor Win % (173,198 runners) | 25.3% | 1.04 |
The favourite was priced almost perfectly. The market expected it to win 34.5% of the time and it won 34.4%. Blind backing of the favourite has no edge in this sample.
The price gradient is flat too. Rank every runner by Betfair SP and the favourite, second, third, fourth, fifth and the long-shot pile from sixth onwards all sit between A/E 0.99 and 1.01. Strike rate falls steeply down the field, and the price compensates almost exactly.
Single ratings don’t break the market either. None of them on their own beat Betfair SP over these 13 years.
Combining ratings raised the strike rate but not the value. When PR Rank 1 and the top Official Rating are the same horse in a handicap, the win rate jumps to 29.0%. That looks transformative. The A/E is still 0.99. Agreement gives you winners, and it doesn’t give you a better price.
Race Predictor is the one signal in this list that moves above 1.00. The market expected its top-ranked horses to win 24.4% of the time and they won 25.3%, an A/E of 1.04. It held at 1.04 across Flat all-weather and Flat turf, 1.03 over hurdles and 1.06 over fences. In this sample that’s an edge of 3% to 6% over Betfair SP.
That’s not a profit figure. It’s measured before commission, it says nothing about the losing runs along the way, and a top-ranked horse still loses most of the time. One possible reason is that the market absorbs single ratings easily but not a probability built from many factors at once. This data doesn’t prove why.
All of this is history. It shows where the market has been fair and where it hasn’t, not where it has to stay that way.
When can you trust an A/E number?
An A/E is a measurement, and measurements have noise. The fewer bets in the sample, the more that noise can look like an edge. A 1.40 across 50 bets is easy to get by luck. The same number across 5,000 bets is harder to wave away.
Sample size is only one check, though. It doesn’t tell you how the selections were chosen or whether the result holds on races it hasn’t seen.
These are my working thresholds, not laws:
| Bets in the sample | How I read an A/E |
|---|---|
| Under 1,000 | Suggestive only |
| 1,000 to 5,000 | Above 1.10 is interesting, so stress-test it by season, race type and class |
| Over 5,000 | Above 1.05 is worth taking seriously, and above 1.10 is strong |
| Below 0.95 on a similar sample | The angle is bleeding value, even if it’s throwing up winners on a good run |
Look at the spread inside the sample, too. Say an A/E of 1.07 breaks down into 1.20 in Class 5 handicaps and 0.95 everywhere else. It looks like a Class 5 angle, not a universal one. But if you only found Class 5 by slicing the data again and again, test it on fresh races before you trust it. A boundary you set beforehand, for a racing reason, is a different matter.
How do you use A/E to test a betting angle?
The routine is the same whether you’re testing a tipster, a tipping service, a magazine system or your own filter:
- Collect every selection at Betfair SP, not at the recommended odds. The price you actually get is the price the market actually offered.
- Convert each price to an implied probability (1 divided by the decimal odds) and add them up. That’s your expected wins.
- Count the actual wins and divide by expected. That’s your A/E.
- Split the sample. Work out A/E by year, race type, class and going. A real angle should usually hold across most of the slices you’d expect it to. If it only lives in one window of time and you found that window by digging, the pattern has probably moved on.
- Ask whether it makes racing sense. An explanation is a useful sanity check, not proof, because a story is easy to invent afterwards. And a real mispricing won’t always have a neat one.
Then test it on data it hasn’t seen. How to test a system before you bet it live takes you through that, and why horse racing betting systems fail shows what happens when you skip it.
What this doesn’t tell you
A/E tells you whether the prices were fair across a group of bets. It doesn’t tell you what will win tomorrow, and a good A/E in the past can fade. The figures above are UK racing from July 2013 to March 2026, priced at Betfair SP, before commission.
If you want to see the measurement applied to today’s runners, the Race Predictor shows a horse’s probability and its Betfair-implied price side by side, so you can see whether each runner looks overpriced, underpriced or fairly priced. See today’s Race Predictor probabilities.
