Don’t rely on Dosage for Kentucky Derby handicapping

April 22nd, 2019

While no longer as much in the spotlight, the modern Dosage system of pedigree shorthand still gets trotted out in the build-up to the Kentucky Derby (G1). It’s understandable that horseplayers, and fans uninitiated into the mysteries of Thoroughbred bloodlines, want to latch onto a figure that points to a horse’s ability to handle 1 1/4 miles on Derby Day. But please don’t.

Before explaining why, a preamble: I’m limiting this discussion to the Dosage Index (DI) as developed by Dr. Steven Roman, and applied as a handicapping tool for the American classics since the 1980s. Although the deeper history of Dosage goes well beyond the scope of a blog post, it’s vital to know that Roman was modifying a theory pioneered in Europe by Lt. Col. J.J. Vuillier and further adapted by Franco Varola. (If you’re interested in a simplified historical background, and Roman’s approach, see the postscript.)

As a pedigree enthusiast, my immediate problems with Dosage as a Derby prognosticator are the stallions who count (or not) and the deliberate exclusion of any mares from the equation. Its fundamental problem is a mechanistic approach that fails to reflect the complexities of inheritance. Since that scientific critique’s best left to pedigree professionals and geneticists, I’m concentrating on the first two. These observations are hardly new or unique, but worth highlighting as we look to the 2019 classics.

Dosage is supposed to recognize the most statistically influential sires in a pedigree, but Roman ceased to update this list of chefs-de-race upon which his formulas depend. Indeed, three of the last four Derby winners have had Dosage Indices in excess of the magic 4.0 cutoff – Triple Crown champion American Pharoah (4.33), Nyquist (7.00), and Always Dreaming (5.0). The old sires are fading beyond the fourth generation, the limit of Roman’s calculations, without new faces to keep the chefs-de-race current.

But based on Roman’s curious methodology, the list has been problematic for quite a while. He already didn’t see fit to include titans like Storm Cat or Danehill (based on rather forced reasoning) while adding a few sires of lesser merit that haven’t stood the test of time. Among the significant omissions are Sunday Silence, Distorted Humor, Seeking the Gold, Deputy Minister, Unbridled’s Song…I could go on.

So as you can see, the Dosage Index is based on an aged selection of sires that excludes some of the leading progenitors of recent years – the very ones we’d look to for classicity.

Yet even if it were being continuously updated, the very fact that stallions would arise on the scene to become chefs-de-race shows the limits of Dosage as a predictive tool for handicapping. The numbers were always going to be subject to revision based upon new data. A Dosage that’s “too high” one day can be within the parameters the next.

A memorable case in point is 1991 Derby winner Strike the Gold: he entered the starting gate with a 9.0 Dosage Index. But his sire was Alydar, the near-misser to Affirmed in all three legs of the Triple Crown. Roman fixed that by making Alydar a “Classic” chef-de-race, bringing Strike the Gold’s DI down retroactively to 2.60.

Still, Strike the Gold’s broodmare sire, Hatchet Man, never counted at all. Not that he has claims to be a chef-de-race, but Hatchet Man figured to add stoutness if you’re just trying to gauge the distance capacity of an individual runner. A slavish adherence to Dosage, however, would hack Hatchet Man right out of the equation because he’s not elite enough to have a statistically determinable influence.

And mares don’t count at all as progenitors, based upon their limited number of offspring compared to stallions. But again, it’s unwise to airbrush half of the ancestry out.

One glaring example is Crème Fraiche, whose Dosage Index was astronomical when he won the 1985 Belmont (G1). Roman’s questionable solution was to make his paternal grandsire Crème dela Crème a “Classic/Solid” chef-de-race to bring the numbers into line. I’d rather look to Crème Fraiche’s dam, the prolific racemare Likely Exchange, winner of the 1979 Delaware H. (G1). (Her own broodmare sire, Swaps, doesn’t count in Dosage either.). Horses don’t automatically produce offspring who can stay at least as far as they do on the racetrack, but Likely Exchange descends from the potent female line of Escutcheon, ancestress of Shuvee among others. That matrilineal factor simply isn’t accounted for in Dosage, yet it’s arguably why both Likely Exchange and Crème Fraiche emerged from otherwise forgettable sires.

Real Quiet (DI 5.33), who came within an eyelash of the 1998 Triple Crown, is another whose family must be mentioned in any discussion of his classic performances. His second dam was a full sister to 1969 Derby and Preakness winner Majestic Prince. That actually rendered his Dosage higher because the mare’s sire Raise a Native and broodmare sire *Royal Charger count as speed influences in the “Brilliant” category.

Similar issues must be borne in mind when considering the Dosage Indices for the 2019 Kentucky Derby contenders.

Win Win Win (DI of 4.50) summarizes the problems inherent in a Dosage calculation where Lost Code (?!) counts as a chef-de-race and Sunday Silence doesn’t. Moreover, ancestor Unbridled is credited under the “Brilliant/Intermediate” categories – not Classic – as if he were pushing the dial more in the direction of speed. That strikes me as inherently illogical for a stallion of his import on the American classic scene. Unbridled’s classification is also contributing to the over-the-threshold DIs of Long Range Toddy (4.33) and Gray Magician (5.0) (both of whom descend from the uncounted Unbridled’s Song).

Also exceeding the theoretical limit is Improbable (4.23), despite being out of an A.P. Indy mare. Even if you don’t want to give Improbable’s sire, City Zip, credit for getting Collected and Dayatthespa, I’d invoke his deep female line. Improbable hails from the immediate family of Hard Spun, and tracing further back, his maternal relatives include Darby Dan’s dual classic heroes Chateaugay and Little Current. That’s not to deny the influence of their sires, but a female line so productive over the long haul can’t be discounted.

Finally, notice that horses of such divergent pedigree profiles as By My Standards and Maximum Security have the same DI of 3.0. By conventional standards of pedigree analysis, By My Standards would have more questions to answer at the Derby distance. Even viewed through the prism of Dosage itself, his entire Index leans on a threadbare three chefs-de-race while Maximum Security at least has a few more classic-oriented ancestors on the list.

Reading the pedigree tea leaves is never easy, especially with the overall trend toward more speed, and less redoubtable stamina influences than in the past. Still, relying on a questionably derived number isn’t the solution.

Postscript on the history of Dosage

In the early 20th century, Vuillier undertook laborious work on the pedigrees of classic winners in Europe, discovered the elite stallions (chefs-de-race) who factored repeatedly back to the 12th generation, and quantified their standard amount of representation. His calculations made clear if a stallion or mare’s pedigree lacked the right amounts of certain chefs-de-race, and accordingly helped identify mates who could redress the balance. Vuillier also heeded the extraordinary influence of the 19th-century mare Pocahontas and included her as a key ancestor – unlike the subsequent spinoffs from his theory. Vuillier’s novel Dosage system was put into great effect by the Aga Khan (grandfather of the present one).

The next step came courtesy of Varola, who built upon Vuillier’s concept of chefs-de-race. Instead of just updating the list of prepotent sires, Varola had the inspiration to classify them according to the essential quality transmitted to their offspring, their aptitudes. He began with five groups – Brilliant, Intermediate, Classic, Stout, and Professional – then realized the need to refine further. Not only did some sires warrant placing in two categories, but Varola also began to split up categories to reflect variations even within the same aptitudinal group.

Roman picked up on Varola’s five groups, and in an extreme modification of Vuillier’s method, assigned points to chefs-de-race in a pedigree going back only four generations. He used a sliding scale to allocate the highest points for a first-generation chef-de-race and the fewest for the fourth. The points were arranged in a linear Dosage Profile, spanning the Brilliant through Professional categories (roughly, speed through stamina spectrum), and Roman created accompanying formulas – the Dosage Index (DI) and Center of Distribution (CD) – to sum up whether the pedigree was tilted toward speed or stamina. More details on Roman’s method are available in the old Brisnet library.

For Derby handicapping purposes, a DI of 4.0 was the cutoff, with anything above that believed to be lacking in stamina for the classic distance. Roman’s twist effectively popularized Dosage into a barometer of staying capacity, a direct contradiction of Varola’s intent.

Roman’s adaptation gained currency through the columns of the late Leon Rasmussen in Daily Racing Form. But note that Rasmussen had long been involved in Dosage pre-Roman, having helped Varola in assessing American sires, as Sid Fernando has described in his eponymous blog.

Nyquist wins the 2016 Kentucky Derby (c) Coady Photography/Churchill Downs