A reader once asked me whether you could find winners by adding up the last three finishing positions and backing the lowest totals. It is a question every form student arrives at eventually. The form column is right there on the card, the numbers are tiny and the arithmetic is simple. A horse that went 1, 2, 1 must be doing something the bookmaker has missed… surely.
I tested it when it came up, and again more recently on an earlier analysis of over 250,000 UK runners from 2021 to 2026. This is what came out.
Do the last 3 form figures find winners in horse racing?
They do point towards winners, in the sense that lower totals win more often. Whether that helps you make money is a separate question, because the price you take matters as much as how often the horse wins.
Two quick terms. The strike rate is the percentage of runners that win. Betfair SP is the Betfair Starting Price, the final exchange price at the off.
Why the sum of the last 3 form figures looks so tempting
The pull of this method is the strike rate ladder. In the earlier analysis, I counted only runners with three valid numeric form figures (no fallers, no pulled-up runs, nobody with fewer than three completed races). The relationship between the total and the win rate is almost a perfect staircase.
| Sum of last 3 finishes | Runners | Strike rate |
|---|---|---|
| 3 (a “1, 1, 1” line) | 1,675 | 28.1% |
| 4 | 3,713 | 20.9% |
| 5 to 6 | 14,179 | 19.4% |
| 7 to 8 | 21,313 | 17.4% |
| 11 to 15 | 73,865 | 12.7% |
| 16 to 20 | 60,739 | 9.1% |
| Above 20 | 49,401 | 6.0% |
Those are figures from one stretch of racing, and a different five years would move the decimals around, though probably not the shape of the staircase.
A horse with a “1, 1, 1” line won nearly five times as often as one with a total above 20. So why not back the lowest totals?
Because the strike rate alone does not say whether the price pays enough for it.
What the data says when you test it properly
The standard check is the A/E ratio, which stands for actual over expected: the wins that really happened divided by the wins the prices said to expect. An A/E of 1.00 means a group won as often as its odds implied. If A/E is new to you, what A/E means in horse racing explains it from scratch.
Here is how three signals that are easy to see on a card. They cover every race in Britain and Ireland from 1 July 2013 to 14 October 2026, at Betfair SP, 1 point level stakes (the same stake each time), after 2% commission, with fallers and pulled-up runners counted as losers. The Return column is the profit or loss as a percentage of total stakes.
| Signal | Races | Win rate | A/E | Return | Longest losing run |
|---|---|---|---|---|---|
| Betfair SP favourite | 167,257 | 34.4% | 1.00 | -1.4% | 28 |
| Our VDW Rank 1 (the horse one of our rating systems puts top in each race) | 153,920 | 21.6% | 0.99 | -3.8% | 59 |
| Our Race Predictor top horse (our model) | 167,314 | 25.5% | 1.01 | 0.0% | 36 |
These describe that particular stretch of racing and will not repeat exactly, so read them as a guide to the shape of things.
What they show is that signals with a very respectable strike rate still land close to an A/E of 1.00. The favourite wins about a third of the time and returns roughly what you stake, less a little. A high win rate on its own is not an edge. The same applies to the ladder above: a group winning 28.1% of the time tells you nothing about profit until you know the prices it started at.
Where the form-figure sum might add information
Useful patterns rarely sit in the headline number. They tend to sit in the cross-tab, which means splitting one group by a second factor. Two splits from the earlier analysis are worth showing.
The first split is by code. Among runners with a total of 4 or less:
- Turf chasers won 29.0% across 917 runners.
- Turf hurdlers won 24.6% across 1,393 runners.
The second is by recency, using days since last ran. Among runners with a total of 4 or less:
- Back within a week: 34.2% across 438 runners.
- A gap of 29 to 60 days: 24.5% across 1,676 runners.
- A gap of 90 days or more: 19.9% across 146 runners.
Those win rates do not tell you which group was the better bet at the price.
A better question than “what are the form figures?”
Treat the form-figure sum as a small confirming input rather than a primary filter. A multi-factor model gives a useful comparison. Our Race Predictor puts a win percentage on every runner using ratings, pace, going suitability, course and distance form, and trainer and jockey patterns. Its top horse is also the Betfair favourite about 41% of the time, so the two signals overlap a lot.
In the earlier analysis, runners that the Race Predictor ranked in its top quintile (the top fifth) and that also had a form total of 2 to 4 won 32.6% across 2,699 runners. Top-fifth runners with a total of 11 or more won 19.8% across 28,759. Adding the form total changed the win rate.
The group I find most interesting is the one backers often take on sight. Runners with a recent form total of 2 to 4 but ranked in the bottom fifth by the model won just 7.3% across 355 runners. It does fit with something simple: form figures show what a horse did, not the class of race, the going or the pace it faced.
If you are building a method, the form-figure sum looks more useful for ranking recent results than for finding a price. It may earn a place as confirmation behind a stronger lens. The guide to horse racing systems using form figures goes through how people build on them, and combining ratings covers the idea of layering more than one signal.
Anything you take from here is a rule of thumb, not a tested rule. To test it, write down every qualifying runner before the race, with its Betfair SP and result, and compare the wins you got with the wins those prices implied. The guide on how to test a system live goes through the method. Until a few thousand runners are in the record, treat what you see as suggestive only.
Where to look next
To compare the form column with other things you can see on a card, open today’s race cards and read the form figures alongside the going and the pace.
