Does it still exist ?
With the advent of Betfair and off-course Corporate Bookies offering all sorts of betting products the pricing of horses by the time race starts is so much more accurate than what it was 30 years ago. So does this mean that the best odds available on a horse is much closer to the real price that the horse should be ? In other words does Favourite-Longshot Bias still exist ?
My analysis centred on a rigorous scientific method drawn from my extensive actuarial experience. The first step was ensuring a complete dataset of races to derive statistically significant conclusions. It’s crucial not to haphazardly select races with out-of-the-norm betting margins and/or results. eg. races with dead-heats, races with betting margins greater than 35% etc. By filtering out “dirty” data, like the metaphorical salmon John West rejects, I focused on races with “clean” data in the period June 3, 2023, to April 13, 2025.
Odds is the median value of the final fixed odds offered by the top 7 corporate bookies ; so is effectively SP Odds . Every race has the SP Odds normalised to 100%. So every horse has both SP Odds and Normalised Odds. The Normalised Odds were then grouped into 11 categories with about the same number of runners in each. The analysis was done using the Normalised Odds categories. But for final report the Average SP Odds of all runners in a Normalised Odds category was calculated. Therefore in some categories e.g. [5] 8.41 to 11.08, the Average SP Odds 8.00 was less than the lowest normalised odds in that category.
So in the finally selected cohort of 32 k races, I’ve used all 316 k runners. I normalised the odds of all runners in a race. The metrics used were Strike Rate and A2E :
- A2E = ( Actual Winners divided by Expected winners) minus 1
- Strike rate = Actual Winners divided by the number of runners
This means that for every race the expected number of winners (aka 1.00) equals the actual number of winners (aka 1). And it also means that in the total sample of 32,186 races there are 32,186.00 expected winners meaning that the A2E for the entire sample is 0%. This then allows statistically rigorous conclusions to be reached.
So here’s the results …
They say a picture tells a thousand words and so here’s a graph …
So YES the conclusion is that Favourite-Longshot bias is still a thing !