The Ancestors Worth Doubling: New Findings from 669,000 Runners

When we published our first inbreeding analysis earlier this year, the response was immediate. Breeders, agents and bloodstock students wanted more. More ancestors, more depth, more answers about which patriarchs and matriarchs actually reward duplication and which ones don't.

So we went back to the database and rebuilt the analysis from the ground up. The cohort grew from a 41,000-runner sample to a larger population of 669,411 thoroughbred runners born from 2008 onwards. The stakes-winner definition was tightened to strict black-type only, meaning Group 1, Group 2, Group 3 and Listed wins (no Restricted Stakes, no Black Type catch-all). The baseline stakes-winner rate sits at 3.76%.

Against that baseline, we tested 32 ancestors that the original article didn't cover. Some of the findings confirmed what experienced breeders already suspect. Some of them genuinely surprised us. And one foundation mare myth turned out to be exactly that, a myth.

Three new positive signals worth knowing

Three ancestors stood out as positive contributors when carried on both sides of the pedigree, all on sample sizes large enough to take seriously.

Roberto is the headline finding. Horses with Roberto on both the sire and dam side returned 4.42% stakes winners across 16,710 runners, against 4.08% for runners carrying him on one side only. That's an 8% uplift on a very large sample.

Rough Shod (II), the great GBR-bred mare foaled in 1944, produced an even stronger signal at +11%. Her both-sides cohort of 24,825 runners returned 4.64% SW versus 4.18% for one-side-only horses. She remains one of the most influential foundation mares in the breed, and the data here reflects it directly.

Somethingroyal, dam of Secretariat and Sir Gaylord, returned 4.06% on a remarkable 121,080 both-sides runners against 3.81% for one-side-only. The delta is smaller in percentage terms, but the sample size is enormous, which gives us high confidence that the effect is real rather than statistical noise.

These are the three ancestors a thoughtful breeder should be willing to inbreed to with confidence based on this dataset.

The Roberto sons who drive the signal

The Roberto finding deserves a closer look because it tells us something practical. The +8% uplift across all 16,710 both-sides Roberto runners is interesting in aggregate, but Roberto's sire-line legacy is not uniform. He left several sons who became important sires themselves, and the inbreeding effect varies enormously depending on which of those sons appears in the pedigree.

We isolated the both-sides Roberto cohort and asked: of those 16,710 runners, which of Roberto's sons appear in their pedigrees, and how does each subset perform?

The results are striking.

Dynaformer runners (1,197 horses with Roberto on both sides who also carry Dynaformer) returned 4.93%. Lear Fan runners (2,760 of them) returned 4.93%. Kris S. runners, the largest subset at 4,378 horses, returned 4.64%. All three are well above the baseline of 3.76%, and all three sit above the Roberto both-sides average of 4.42%.

At the other end, the picture is much less encouraging. Brian's Time runners returned just 1.40% on 857 horses. Real Shadai runners returned 1.47% on 543 horses. Red Ransom came in at 3.25% and Silver Hawk at 3.40%, both below the population baseline.

What this tells us is that the Roberto inbreeding signal is really a Kris S./Lear Fan/Dynaformer signal. If you're considering a mating that doubles up on Roberto and the lines run through one of those three sons, the data supports the cross. If the Roberto duplication runs through Brian's Time or Real Shadai, you should probably look elsewhere.

This kind of granularity is exactly the sort of thing that can be checked in the pedigree of any horse using Pedigree Search in G1 Goldmine. You can see exactly which Roberto descendant lines a horse traces to and on which side they appear, and combine that with the findings here to make a more informed assessment.

Rough Shod's modern legacy

Rough Shod (II)'s +11% signal is genuine but it deserves a contextual note. Her direct influence on modern pedigrees flows mainly through her granddaughter Special, dam of Nureyev, and through Special's wider family which produced Sadler's Wells, Caerleon and Fairy Bridge. In other words, when you see Rough Shod duplicated in a 7-generation pedigree of a 2008+ runner, you're often really looking at duplication of the Nureyev/Sadler's Wells/Caerleon family.

The Rough Shod cross-distance data backs this up. The strongest signals appear at depth: 6x6 returns 4.86%, 6x7 returns 4.95%, 7x7 returns 4.71%, all on substantial samples. These are not tight crosses. They are deep, layered ancestral influences flowing through a tightly bunched group of elite descendants.

For breeders, the practical reading is this: Rough Shod duplication is essentially shorthand for "this horse traces multiple times to the Special / Nureyev / Sadler's Wells female line", and that pattern correlates positively with stakes performance in the modern population.

The foundation mare myth: La Troienne and Plucky Liege

Now for the more uncomfortable finding.

La Troienne is widely cited as one of the most influential matriarchs in the breed. Plucky Liege similarly. Both names carry a kind of mythical weight in pedigree circles. So we tested both of them, expecting to see a positive inbreeding signal.

The data does not support the claim, at least not for inbreeding within seven generations.

La Troienne: 3,619 both-sides runners, 2.40% SW. That's 27% below the one-side-only rate of 3.27%.

Plucky Liege: 1,479 both-sides runners, 2.10% SW. That's 30% below the one-side-only rate of 3.00%.

But here's the important nuance. When we looked at the cross-distance breakdown, virtually all of those La Troienne and Plucky Liege duplications sit at 7x7. Both mares were foaled before 1930 (La Troienne in 1926, Plucky Liege in 1912). For a horse born in 2008 or later, having either mare appear at generation seven on both sides is essentially a statistical inevitability across large parts of the breed, not a deliberate breeding choice.

The honest read is not that "inbreeding to La Troienne is bad". It's that La Troienne is now too far back in modern pedigrees to drive a meaningful inbreeding effect. The duplication is so deep and so widely shared that any signal is essentially diluted to noise. The same applies to Plucky Liege, Frizette, Selene and other early-twentieth-century foundation mares.

This is an important calibration for breeders who use these names as a kind of pedigree shorthand for quality. By 2008, the practical influence of a 7x7 La Troienne cross on actual race performance is negligible. The matriarchs that still drive measurable signal at modern depths are the ones from the 1940s and 1950s, like Rough Shod and Somethingroyal.

A.P. Indy: the surprise negative

Among the modern sires we tested, A.P. Indy produced the most counterintuitive result. He's universally regarded as one of the great sires of his generation, and his line remains commercially important. Yet horses with A.P. Indy on both sides returned just 3.01% SW across 3,720 runners, against 4.08% for one-side-only horses. That's a 26% underperformance.

The cross-distance data is consistent with the headline. At 3x3, A.P. Indy returns 2.87% on 663 runners. At 4x4, just 2.62% on 687 runners. At 4x5, only 0.69% on 145 runners. Only the small-sample 5x3 cross (110 runners at 7.27%) bucks the trend.

This is a finding worth taking seriously. A.P. Indy is recent enough that "both sides" still means relatively close inbreeding (typically 3x3 to 4x5), and at those distances the data clearly says the cross does not work.

There is a generational explanation that matters here, and we'll explore it properly in our next article. For now, the practical implication is: if you're sketching out a mating and both prospective parents trace closely to A.P. Indy, the data says you should think twice.

Putting this to work in Pedigree Search

The real value of an analysis like this is not in the headline numbers. It's in being able to apply them to a specific mating decision, sale catalogue, or yearling shortlist.

In G1 Goldmine, Pedigree Search lets you pull up the seven-generation pedigree of any horse in the database and immediately see which ancestors appear on both sides, at what generation, and through which sons or daughters. You can check whether a Roberto duplication runs through Kris S. or through Brian's Time. You can see whether an A.P. Indy cross sits at 3x3 or at the more forgiving 5x3. You can identify whether the Rough Shod influence comes through Special or through one of the less productive sub-branches.

We are also building an Ancestor Duplication panel that will surface this information automatically. For any horse you look at, the panel will identify the most-duplicated ancestors in the pedigree, their cross distances, the sub-lines through which they descend, and which recognised breeding theories are present. Combined with the findings in this article, this turns ancestor duplication from a manual research project into a one-glance check.

We'll have more on that release shortly.

What's next

We have one more piece of analysis to share, focused on what is probably the most interesting puzzle in this dataset: the case of horses where one-side influence is hugely positive but both-sides influence is hugely negative. Galileo is the cleanest example. So are A.P. Indy, Kingmambo, Last Tycoon, and several other recent elite sires. The pattern is too consistent to ignore, and it tells us something fundamental about how to read modern pedigrees.

That article is next.