The man who counted the Cleveland Bays

Every rare breed has a story about the last few animals. The Cleveland Bay's is well known: by the 1950s, England's oldest established horse breed was down to four purebred stallions.

What is less well known is the second rescue — the quieter one, which happened in this century, and which was carried out largely with a computer.

The problem numbers alone can't show

Counting animals tells you less than you would hope. A breed can be growing in headcount and still dying genetically, because what matters is not how many exist but how much variation is left between them, and how evenly that variation is being passed on.

The Cleveland Bay's studbook was closed to outside blood in 1884 — though "closed" has never meant quite as absolutely closed as it sounds. When the Yorkshire Coach Horse studbook itself later shut, a number of those horses had nowhere left to be registered; being partbred Cleveland Bays, they were brought into the Cleveland Bay purebred book. A Grading Register still operates today, by which a partbred mare with sufficient Cleveland Bay blood can be inspected and registered, and her descendants, bred back to purebred stallions over successive generations, can eventually enter the studbook as purebred. Colts cannot progress by that route; only females. Both are pressure valves on a closed population, and in a breed this small they matter.

The breed then went through mechanisation and two wars, and by the time anyone panicked, the population that remained was not simply small — it was concentrated. When Dr Andy Dell eventually put the whole pedigree through analysis, the result was stark: the breed had once had around 182 founders, animals whose bloodlines were genuinely distinct from one another. Generations of well-meant but uncoordinated breeding had reduced that to the genetic equivalent of about two and a half.

Not two and a half animals. Two and a half founder lines, standing in for 182. Everything else had been quietly lost — not through disaster, but through a long series of individually reasonable decisions about which stallion to use.

Borrowing from the zoos

The tool that changed this came from an unlikely direction.

Since the 1970s, zoos had faced a version of the same problem: small populations of endangered animals, scattered across institutions and continents, needing to be managed as one gene pool rather than many. The response was software — SPARKS, the Single Population Analysis and Records Keeping System, paired with an analysis programme called GENES. Between them they had helped pull species including the Przewalski's horse back from the edge.

Dell's insight was that a native horse breed is, in the relevant sense, exactly the same problem. The animals are privately owned rather than institutionally held, but the mathematics of a closed, dwindling population does not care who holds the headcollar.

Getting there took years. The Cleveland Bay studbook reaches back to pedigrees recorded in 1723 — almost three hundred years and thirty-eight generations of records, compiled by hand, complete with the transcription errors that accumulate over a century of human copying. The system crashed repeatedly. When it finally ran, in 2004, the Cleveland Bay SPARKS database was the largest dataset of its kind in the world.

What breeders actually got

The output was deliberately unglamorous: sheets of paper.

For every breeding animal, the software calculates a mean kinship — a figure describing how closely related that horse is to the whole rest of the breed. Rare bloodlines score low and matter most. Common bloodlines score higher, and still matter, but differently.

From that come the recommendations, translated into a colour-graded banding system a breeder can act on without needing the underlying population genetics. The population is split into four groups:

• Green — the top-ranked matches. The pairings that do the breed most good.
• Yellow — the best of the rest. Defensible, and used by careful breeders where the individual animals justify it.
• Amber — frowned upon.
• Red — an emphatic no. A mating that would actively damage the breed.

The system deliberately also discourages the seemingly clever move of crossing the rarest blood with the most common — which lowers the average impressively in one generation while permanently welding the rare lines to the common ones, leaving the breed with less room to manoeuvre forever after.

This is the origin of a vocabulary you will now hear in any serious Cleveland Bay conversation. Breeders talk in colours because a piece of software taught the whole breed a shared language for a problem that had previously been invisible.

The view from the yard

None of which would matter if breeders had ignored it. So it is worth hearing from one who did not.

Lesley Orange breeds Cleveland Bays at Kenley in Derbyshire with her husband Dennis, and stands the Breed Quality Approved stallion Holmside Applejack. She is unambiguous about where she sits: "I am a strong supporter of the Sparks sheets. Green matings only, but I would consider yellow matings, depending on breeding."

On the two lower bands she is blunter still. Amber, she says, she "would never breed"; red "is absolutely NO as they would be damaging to the breed".

She is equally unambiguous that the system's authority took time to win. "The trustees, until recently, were not very supportive of the Sparks sheets," she says, "so the breed does get orange and worse yet red banded foals being born, much to my frustration."

And then the sentence that ought to be the last word on the subject: "There is no reason to breed red banded foals in this day and age when the genetic diversity is so sensitive."

That is what adoption actually looks like from the inside — not a policy announcement, but individual breeders turning down matings they could have made, for years, while waiting for everyone else to come round.

Did it work?

This is the part that matters, and it is unusually well documented, because Dell went back and measured it.

In 2021, working with colleagues at Nottingham Trent University and the University of Lincoln alongside the Cleveland Bay Horse Society, he published sixteen years of results in the journal Ecology and Evolution. The findings:

• The rate of inbreeding fell from more than 3% per generation to less than 0.5%.
• The effective population size in the UK rose from 20 in 1994 to more than 140 by 2020.

Effective population size is the number that actually describes a breed's genetic health — not the headcount, but the number of animals genuinely contributing distinct genes to the next generation. Twenty is the territory where conservation geneticists start using words like critical. A hundred and forty is a breed with a future.

The researchers' conclusion was that the Cleveland Bay is now in a genetically sustainable position. For a breed that had four stallions left seventy years ago, that is an extraordinary sentence to be able to write.

What it became

The method did not stay with one breed. Under the Rare Breeds Survival Trust, the same approach has been extended to the Suffolk Horse, the Shire, the Eriskay, the Dales and Highland ponies, and to Vaynol cattle. The underlying system has since been developed into a Breed Conservation and Management System — the colour-graded recommendations breeders now work from — and in 2025 it was folded into the Grassroots studbook platform, which is roughly the point at which a specialist rescue tool becomes ordinary infrastructure.

Dell completed his doctorate on the subject at the University of Lincoln in 2010, as an invited mature student — that is, he did the work first and the qualification afterwards. He received the Rare Breeds Survival Trust and Marsh Christian Trust Award for technical support of biodiversity conservation in 2008.

It should be said plainly that no one person saves a breed. The published work is co-authored; the Cleveland Bay Horse Society maintained the studbook that made any of it possible; and none of it would have mattered if breeders like the Oranges had not agreed, one mating at a time, to take the harder recommendation over the convenient one. But there is a reasonable case that a database built in the early 2000s did more for the Cleveland Bay than anything since the society's founding.

The part that still does not work

One thing should be said, because the breeders living with the system say it.

Sparks grades a mating between two animals. But since the sheets moved online, a breeder can generally only see the horses they own themselves. That makes it difficult to look up a prospective mate before agreeing to a mating — which is the one moment the information exists for.

Lesley Orange puts the case plainly. Under the old paper sheets, "you could look up any horse, which was brilliant when searching for more breeding stock". Now, she says, "you cannot pull up an outside mate to see Sparks match (or not)". She hit exactly that wall recently with a mare on breeding loan whose elderly owners were unwell and could not retrieve the data themselves: "I couldn't access her sheets… breeders need to be able to look females up in order to verify Sparks status."

A tool that has demonstrably rescued a breed is worth defending. It is also worth saying that access is now the thing holding it back, and that this is a solvable problem rather than an inherent one.

Why we are writing about this

Because it is the clearest existing answer to a question we get asked often: what would it actually look like to take genetic diversity seriously in a marketplace?

The Cleveland Bay answer is that you make the population-level picture visible to the person making the decision, at the moment they make it, in a form they can act on without a genetics degree. Not a lecture. A colour, on a sheet, next to the animal's name.

That is the standard we have in mind for our own rare and heritage breeds section, and it is why Applejack's listing says green matings only rather than quietly leaving it out. It is also why, when a breeder tells us a stallion is not available to outside mares, we publish that as it stands. A breed managed this carefully is not served by a marketplace implying availability that doesn't exist.

Twenty years ago someone decided that the way to save a horse breed was to spend several years wrestling with corrupted pedigree records until the software stopped crashing. Every breed with a future owes something to that kind of patience — and to the breeders who then did what the sheets told them.

---

Further reading

• Dell, A. et al., "16 years of breed management brings substantial improvement in population genetics of the endangered Cleveland Bay Horse", Ecology and Evolution, 2021 (open access)
• Nottingham Trent University: "Scientists help safeguard genetic diversity of England's oldest established horse breed"
• Dr Andy Dell's research profile
• The Cleveland Bay Horse Society
• Rare Breeds Survival Trust

---

Corrections, 30 July 2026. This piece originally described the Sparks banding as a two-way choice between green and red matings; it is in fact four bands — green, yellow, amber and red — and has been corrected. It also stated without qualification that the studbook closed to outside blood in 1884; a paragraph has been added on the later admission of Yorkshire Coach Horses and on the Grading Register, both of which qualify that. Kenley is in Derbyshire, not Nottinghamshire as first published. Our thanks to Lesley and Dennis Orange for all three.

Lesley Orange is quoted with her permission.

— Rene