Breeding Theories and Statistics

Occasionally, and through various outlets, we publish articles that try to place modern Dosage methodology in its proper perspective.  Invariably we caution against unwarranted extension of its basic premises.  For example, the confirmed relationship between pedigree type and performance applies technically only to large populations of runners and not necessarily to individuals, although the conclusions derived from the data may be considered as would any conclusions drawn from statistical studies.  For example, five-pack-a-day smokers generally don't live as long as non-smokers.  Nevertheless, there are undoubtedly some that live well beyond the age that statistics suggest.  On the other hand, refraining from smoking does increase the probability of longer-term survival.  One should not conclude that every heavy smoker dies young or that every non-smoker lives a very long life even though the statistics tell us that heavy smokers as a group tend to die earlier than do non-smokers as a group.

We describe Dosage as a technique for classifying Thoroughbred pedigree type in terms of inherited prepotent qualities of speed and endurance.  Research through the years has unequivocally demonstrated that the pedigree type of successful performers in any particular performance category, i.e., by distance, age, racing surface, class of race, etc, can be statistically differentiated from the pedigree type of successful performers in another performance category.  Furthermore, in virtually all cases these differences can be correlated with the average distance of the races in the respective categories.  Over the years, however, this simple concept has been extended, modified, and manipulated by some to justify their own perceptions of racing reality.  This has been done without malicious intent in most cases, but rather reflects a general human trait that often leads us to oversimplify a problem that is, in fact, extraordinarily complex.  This desire for simplicity is understandable and seems to occur whenever some theory about racing or breeding is offered up for discussion.  Whether it be Dosage, nicking, inbreeding, LH-X or a multitude of other theories, its application is often naive and misguided, leading to disappointment and ultimate dismissal of the theory's valid points.  In this article we will present our views about the currently fashionable theory of nicking, an idea that seems attractive on the surface but upon critical examination has, with some exceptions, a marginal scientific basis and is subject to gross misuse when accepted at face value. 

The concept of two bloodlines or individuals having a unique compatibility is quite appealing because the decision-making process is that much easier.  The fundamental problem with individual nicking patterns, however, is that only rarely are there enough examples to provide a statistically significant sample size.  By statistically significant we are referring to the established statistical criteria affording a meaningful confidence level where the observed pattern has only a small probability of being a random event.  For example, suppose sire A is bred to mares by sire B and ten foals are produced.  If two of these, or 20%, become listed stakes winners and sire A gets 10% overall stakes winners, i.e., regardless of his mares' bloodlines, then this A/B cross surpasses sire A's overall stakes winner production by two to one, which some would consider evidence of a positive nick.  On the other hand, suppose that none of the ten foals (0%) won any stakes but three placed in Grade I's.  zero stakes winners from ten foals is a red flag.  But here the truth may be that the Grade I stakes-placed runners are far superior to the two listed stakes winners in the first case.  Do 20% stakes winners represent a positive nick while 0% stakes winners represent a negative nick?  Probably not.  Actually, depending on how you measure the quality of runners (and we all do it differently), you might conclude that the second case indicates a greater affinity between sires A and B than does the first.  In extreme cases such as five Grade I stakes winners from five foals produced by the A/B cross or, at the other end, five maiden claimers, the conclusions are more obvious.  But these circumstances are so rare that to generalize about the validity of individual nicking patterns on this basis stretches credibility.  Ten or even twenty or thirty specific examples of a cross between A and B may still not be sufficient to establish statistical significance simply because there is no objective universal measure of the success of that cross. 

An example of the danger inherent in the use of small sample sizes for defining nicks is the breeding of Baldski to Diplomat Way mares.   In 1989, The Blood-Horse published tables of nicking patterns for many North American sires.  Included were the records of sires which had gotten at least five foals from daughters of a particular broodmare sire.  At that time the Baldski/Diplomat Way pattern had resulted in no stakes winners from 19 foals, seemingly supporting the notion of an inferior cross.  However, the AEI (Average Earnings Index) for these foals was 1.84 while the CI (Comparable Index) was 1.64.  So although there were no stakes winners, Baldski did upgrade the Diplomat Way mares to which he was bred.  Perhaps the situation wasn't so bad after all.  Subsequently, there appeared a Baldski/Diplomat Way stakes winner named Express Star which ultimately won seven stakes races with lifetime earnings of over $450,000.  One out of twenty, or five percent stakes winners was not statistically different from Baldski's 7% lifetime stakes winner production rate at the time.  Suddenly the prospect of Baldski being bred to Diplomat Way mares wasn't so terrifying.  As of October 2000 there were 28 foals representing the Baldski/Diplomat way cross.  Of these, three (11%) are stakes winners, exceeding Baldski's 9% lifetime percentage of stakes winners.  The combined AEI of the 28 foals is 2.03 (which exceeds Baldski's lifetime AEI of 1.77 through 1999), with a CI of 1.62.  

The issue raised here relates more to timing than it does to any genetic compatibility.  Does a nick change if the one stakes winner from 20 foals is the first foal or the last one?  Obviously not.  The total record after 20 foals is the same.  Bloodline compatibilities can change over long periods of time as new breeding stock from those bloodlines emerge.  But the idea that the compatibility between a specific sire and a specific broodmare sire can change over time is suspect. 

Another problem with the small sample sizes used in defining individual nicking patterns is the tendency to revert to close up ancestors if there are no specific examples involving the particular sire and broodmare sire.  This is more often the case than not.  For example, millionaire Dispersal was one of the very best colts in America several years ago and presumably evidence of  a superior nick.  Dispersal was by Sunny's Halo and out of a Johnny Appleseed mare, certainly not a well-established breeding pattern and for which there is no basis for drawing meaningful conclusions about the viability of the cross.  Even today there are only three examples, and Dispersal remains the one stakes winner.  If we now look at the record of Sunny's Halo's sire, the deceased Halo, who was a superior stallion of long-standing, we find that he has sired at least five foals from mares representing only one branch of the Prince Rose broodmare sire line (like Johnny Appleseed).  From six  Prince John mares he has gotten one minor stakes winner, but overall he has seriously downgraded these mares with AEI 0.93 compared to CI 1.97.  This is not the stuff that great nicks are made of.  Apparently, using any other representative of a bloodline as a basis for the compatibility of another member of that bloodline is a dangerous practice because with each succeeding generation there is a significant dilution of the genetic relationship between the two. 

One can look at nicking patterns more globally by focusing only on bloodlines rather than individuals.  For example, it can easily be determined what percentage of Nearctic line sires and Royal Charger line broodmare sires are present in a population of stakes winners.  For argument's sake let's say that 20% of all stakes winners are by Nearctic line sires and that 10% are out of Royal Charger line mares.  In a random world, then, we could expect that one in ten stakes winners sired by a Nearctic line stallion would be produced from a Royal Charger line mare.  Conversely, one in five stakes winners produced from Royal Charger line mares would be sired by Nearctic line stallions.  To avoid confusion, we'll use real numbers to illustrate.  In a population of 100 stakes winners, a total of 20 would represent the Nearctic sire line.  If, randomly, 10% of all stakes winners are from Royal Charger line mares then two of the twenty Nearctic sire line stakes winners would be expected to represent the Nearctic/Royal Charger cross.  Similarly, since 20% of all stakes winners are by Nearctic line sires then two of the ten stakes winners from Royal Charger line mares would be by those sires.  Thus, from either direction, we expect two out of 100 stakes winners in the population to represent the Nearctic/Royal Charger cross.  If, however, we find that there are actually four or six of the hundred that are bred Nearctic/Royal Charger, then we have a situation in which two or three times as many representatives of that cross have been produced than had been anticipated from the total population statistics.  This might be construed as a positive nick.  But here, too, there are problems associated with interpreting the data. 

First, the derived statistics and nicking patterns, as is the case with Dosage, would apply only to large populations and not necessarily to individuals.  In fact, we breed individuals, not bloodlines.  Second, do the statistics really reflect general compatibility of bloodlines or the quality of individuals that make up the population?  Bloodlines rise and fall for many reasons.  But a characteristic of ascending bloodlines is that the individuals representing that bloodline are superior breeding animals relative to the remainder of the population.  Declining bloodlines are similarly characterized by inferior breeding animals.  Therefore, dominating the situation is the fact that breeding superior sires to mares by superior broodmare sires increases the probability of producing superior foals.  This returns us to the concept of individual nicking patterns and the problems associated with statistical significance. 

Bloodline compatibilities are statistically meaningful only in terms of large populations and in that context are useful in assessing general trends within bloodlines.  On the other hand, individual nicking patterns almost never meet the criteria for statistical significance.  

The overall concept of nicking must be placed in perspective.  It can not and should not replace a careful analysis of the quality of the breeding stock being evaluated.  Good breeding stallions get good foals.  Bad breeding stallions do not.  Good mares produce good foals.  Bad mares do not.  It isn't an accident that typically just 1% of all the sires of starters in North America account for about 25 to 30% of  all stakes wins and available purses.  With that in mind, bloodlines are of secondary importance when compared to individual sires and dams.  The stallion and the mare must be considered in terms of their individual attributes. 

Because individual nicking patterns in the vast majority of instances are statistically insignificant and because the general compatibility of bloodlines ignores individual contribution, our view is that the best way to proceed in selecting a mating is to identify the most productive sires that are economically consistent with a particular mare and, for comfort's sake if nothing else, look for an indication that either the sire has gotten at least one or two good runners from the mare's sire line or, conversely, that the mare has already produced at least one good runner from the sire's sire line.  The breeder can then invoke secondary analyses relating to potential commercial value of the offspring or, if racing rather than sales is the goal, racing type (including distance potential, surface preference, rate of maturity, etc). 

Equally important is a thorough understanding of the physical compatibility of the stallion and the mare, especially with regard to conformational attributes and the degree to which these conformational characteristics have effected the performance of the progeny produced.  Some form of nicking may serve as a preliminary focal point, but in the end it should never be the basis for selection.  Barrera is not Mr. Prospector, although both are sons of Raise a Native.  Wajima is not Secretariat, although both are sons of Bold Ruler.  Anyone who believes a Barrera/Wajima foal represents the same nicking significance as a Mr. Prospector/Secretariat foal had better go back to basics.