Every racing pundit will tell you a horse "wants soft ground" or "needs it quick". We have a database of UK and Irish results, so we asked the harder question: does the ground genuinely split a sire's record, and if it does, can you make money from it?
Pedigree is meant to be the first clue to how a young horse will handle the ground, before it has form of its own. We took every settled runner in our database from January 2018 to June 2026 that had a known sire and a recorded going, bucketed the ground into three honest groups (soft for soft, heavy and yielding; fast for good and firm; all-weather for the synthetic surfaces), and measured each sire's strike rate, A/E and return at starting price on each. That is 914,831 runners.
Before naming names, the sobering part. Across all sired runners, strike rates and returns barely move between the three ground types, and every group loses to level stakes at SP. The market has already read the form book.
| Ground | Runners | Win rate | A/E | ROI at SP |
|---|---|---|---|---|
| Fast (good, firm) | 337,308 | 11.6% | 0.89 | -20.6% |
| Soft (soft, heavy, yielding) | 342,916 | 11.2% | 0.90 | -21.6% |
| All-weather | 234,523 | 10.5% | 0.85 | -25.5% |
A/E is actual winners divided by market-expected winners (the sum of every runner's implied chance). Below 1 means a group won less often than its prices implied. All three sit below 1, which is the bookmaker's margin doing its job.
The interesting signal is not the average, it is the spread between a sire's two records. Some stallions pass on stamina and a love of cut in the ground; others pass on speed that wants a sound surface. Here are the sires whose progeny show the sharpest preference for fast ground (the negative swing) and for soft (the positive swing), among those with a decent sample on both.
| Sire | Soft win% | Fast win% | Soft minus fast |
|---|---|---|---|
| Motivator (GB) | 18.0% | 12.5% | +5.5 (soft) |
| Territories (IRE) | 15.3% | 10.3% | +5.0 (soft) |
| Fast Company (IRE) | 11.2% | 8.1% | +3.1 (soft) |
| Dansili (GB) | 7.5% | 12.4% | -4.8 (fast) |
| War Command (USA) | 5.9% | 12.2% | -6.2 (fast) |
| Robin Des Champs (FR) | 11.1% | 17.5% | -6.4 (fast) |
| Paco Boy (IRE) | 4.5% | 11.1% | -6.5 (fast) |
These are real, sizeable splits, not noise: War Command's runners win more than twice as often on fast ground as on soft, and the swing holds across hundreds of runs each side. Used as a profiling tool, that is genuinely useful, especially for a horse running on a surface it has never tried, where the sire is the only ground clue you have.
If you want the names that win most often when the mud is flying, the list is dominated by jumps and staying sires, exactly as you would expect. Strike rate alone, with a minimum of 400 runs on soft:
| Sire | Runs (soft) | Win rate | A/E | ROI at SP |
|---|---|---|---|---|
| Doctor Dino (FR) | 589 | 20.5% | 0.98 | +5.2% |
| Masked Marvel (GB) | 409 | 20.1% | 1.09 | +3.4% |
| Saddler Maker (IRE) | 854 | 19.0% | 1.01 | -6.9% |
| Great Pretender (IRE) | 819 | 18.1% | 1.06 | +7.1% |
| Shantou (USA) | 2,720 | 16.1% | 1.09 | +5.4% |
A few of these clear break-even on soft, and Shantou does it across a meaty 2,720-run sample with an A/E above 1, which is rarer than it sounds. Treat the smaller samples (Doctor Dino, Masked Marvel) as promising rather than proven: a few hundred runs is enough to be interesting, not enough to bank on.
On a sound surface the names shift towards classier Flat lines, plus a couple of jumps sires whose stock clearly wants it quick:
| Sire | Runs (fast) | Win rate | A/E | ROI at SP |
|---|---|---|---|---|
| Kings Theatre (IRE) | 768 | 19.9% | 1.19 | +13.5% |
| Dubawi (IRE) | 2,045 | 19.4% | 0.90 | -15.8% |
| Frankel (GB) | 2,104 | 19.0% | 0.96 | -7.8% |
| Robin Des Champs (FR) | 510 | 17.5% | 1.05 | +21.6% |
| New Bay (GB) | 1,009 | 16.0% | 1.01 | +7.8% |
Here is the lesson in one row: Dubawi's runners win nearly one fast-ground race in five, the second-best strike rate on the list, and still lose you 15.8% at SP. A famous, fashionable sire is short in the market for the same reason a short favourite is: everyone can see it. Kings Theatre is the opposite, a less glamorous name whose fast-ground runners have outperformed their prices (A/E 1.19), though again the 768-run sample asks for caution.
Two things, and they pull in different directions. First, ground preference in a pedigree is real and measurable: the soft-versus-fast splits above are too large and too consistent to be chance, and they are most valuable for a horse trying a surface for the first time, when its own form cannot help you. Second, knowing a sire likes the ground is not the same as having a bet: on the headline names the market has already shortened the price, and strike rate keeps marching ahead of profit.
So use sires by going the way we do, as one input that helps read a race, not as a system to back blind. The edge is never "this sire likes soft". The edge, if there is one, is a horse whose ground suitability the market has under-priced. Most of the time there is no such gap, and the honest answer is to leave the race alone.
The Overlay is a free, data-driven UK and Irish racing model. Every selection is logged before the race runs and settled in public, win or lose. It is not a tipster, and it tells you when to sit a race out. Free value board, bet calculator and a play-money tracker.
Explore The OverlayWe used every runner in our database from 1 January 2018 to 22 June 2026 with a settled finishing position, a recorded sire, a starting price above 1.0, and a recorded going: 914,831 runners. Ground was bucketed by the going string into soft (soft, heavy, yielding), fast (good, firm) and all-weather (standard and synthetic surfaces). Win rate is wins divided by runs. A/E is wins divided by the sum of each runner's implied chance (1 divided by decimal SP). ROI is level-stakes return at starting price: stake 1 point on every qualifying runner and measure profit or loss per point. Strike-rate tables use a minimum of 400 runs on the relevant ground; the split table a minimum of 250 runs on each of soft and fast. Figures are historical, sample-dependent, and for information only.