378 Making-Do When Big Data Isn’t Big Enough: Limitations and pathways of possible development in the prediction of Thoroughbred racing performance as a breeding outcome

Abstract

Abstract A performance genetics model adapted to real-world data of Thoroughbred racing and breeding will be presented. The continuity of ancestry and performance is documented in the annals of pedigree, not as functional counterparts, but as corresponding signs of market value. Pedigree invites chronic misprision of highly diverse genetic resources competing for scarce opportunity in the racing economy, resulting in samples that are often small and of dubious composition. Data is specialized for competing “pedigree analysis” that packages the system of signification to meet market demand for meaning. Given statistical deficits, this data is a necessary inferential asset of the model. The model’s premise is that predictability is optimized if racing performance is defined as the function of an indivisible relation between parents. Statistical data consists of 6-generation ancestries of mares that produced offspring by a subject stallion. Comparison of proportions is used to identify effects resulting from his relation to individual ancestors of the subject mare. Expected performance by the sire’s offspring is defined as the proportion of mares that produced a superior runner by him. Each ancestor of a subject mare also has descendants among the mares that produced offspring by the stallion. For each of those groups, the proportion of mares that produced a superior runner is compared with the stallion’s expected performance using a t-test of statistical significance at the .10 level. Probable effect is further tested by case study involving such variables as racing class, generational distance, sex-linkage, inbreeding, and an ancestor’s pattern of effect across the stallion population. Stallions with the highest prevalence of positive effects are preferred for the subject mare. This model, under the trade name LyonScore®, has been used since 2012 by Werk Thoroughbred Constultants, Inc. as a component of its client services. Table 1: The data, listed in tabular format below, is graphically displayed for actual use on an ancestry tree whose nodes are numbered by relation to a subject mare (“Position”). Each statistical data item in the table is derived from the stud record of a stallion named Distorted Humor and corresponds with an ancestor of a mare named Positively Royal. Each ancestor of this mare is also an ancestor of a group of mares that produced offspring of Distorted Humor and were at least three years of age as of 2019. A proportion of each of those groups of mares produced at least one superior runner. Proportions that differ significantly from Distorted Humor’s expected proportion are so indicated. Only ancestors involving a group of at least 18 mares are considered to have inferential value on statistical grounds. Table 2: The sire Danzig is an ancestor of Positively Royal, along with 75 mares that produced foals by Distorted Humor, only five of which produced a superior runner by him, significantly fewer than expected. However, since Distorted Humor’s dam is by Danzig, the question of generational distance is relevant as a variable to the effect of inbreeding. The table below shows Danzig’s proportional distribution by genetic relation to those 75 mares as indicated by “Position.” Of the 64 mares in descent of Danzig within three generations, three mares produced a superior runner. It’s notable, though, that 2 of 6 mares with the same relation to Danzig as Positively Royal produced superior runners. Since Danzig is the only ancestor with a negative effect, further consideration is warranted. Table 3: In a population that tends to slough off unprofitable genetic resources, overspecialization is the main risk of close inbreeding. Some generational variations of an ancestor’s contribution can turn inbreeding to less specialized effect, but this depends on generational distance. Distorted Humor’s earliest opportunity with mares in descent of Danzig involved offspring closely inbred to Danzig, but later in his career he encountered mares for which variation was more likely. As the table shows, two of six mares descending from Danzig in position 25 (4th generation) produced Distorted Humor’s best runners inbred to Danzig, so it is of some interest that Positively Royal, the subject mare, is also a postion-25 descendant of Danzig. However, that those two mares happen to be full sisters whose breeding has little else in common with that of Positively Royal leaves the question of Distorted Humor’s fitness for this mare less certain than would be preferred.