The Unenviable Role of the Egghead

published: 31 Aug 2015 in Pedigree analysis

In my considered opinion, eggheads get a bad rap. Some would say it’s not surprising I would want to bring that up for consideration. Yes, the shape of my head has been mentioned on occasion, but, like our President Obama when his signature Affordable Care Act was ridiculed as “Obamacare,” I’ve learned to embrace it. The credit goes to Sid Fernando, who would remind me about it from time to time after he took over as CEO of Werk Thoroughbred Consultants. We joke about it now. 

In my considered opinion, eggheads get a bad rap. Some would say it’s not surprising I would want to bring that up for consideration. Yes, the shape of my head has been mentioned on occasion, but, like our President Obama when his signature Affordable Care Act was ridiculed as “Obamacare,” I’ve learned to embrace it. The credit goes to Sid Fernando, who would remind me about it from time to time after he took over as CEO of Werk Thoroughbred Consultants. We joke about it now. 

   Nobody would accuse Bill Oppenheim of being an egghead, but he’s good at dreaming up questions that are murky enough to require one. Not every egghead is as unabashed about it as I am, so I apologize in advance to Prof. Emily Plant, with whom Bill is associated in their SSR Group. When it comes to splitting variables into discrete effects and measuring how they interact, Emily gets the job done. For every question Bill can dream up, Emily knows how to find the answer.

  Emily and Bill are breaking new ground in the analysis and representation of pedigree quality, especially in regard to sire performance. They have parlayed Bill’s familiar APEX rating into the context of a much larger analysis that takes into account the quality of the mares covered by a sire from year to year. Using these measures, they graph the sire’s performance from crop to crop in relation to the quality of his mares and project the performance of crops that have not yet raced. They have also invented a statistical mechanism for  predicting the potential of new stallions, and most recently they began exploring the role speed figures might take in forecasting the ability of young sires.

  Pedigree quality is the main challenge to pedigree research, whether a preoccupation, as it is to Bill and Emily, or an inconvenience, as it is to me. My oval head, whether because of its circumstances or its nature, is fascinated with pedigree relations. Pedigree quality gets in the way. After all, for any given pedigree relation, how do I know I’m measuring the effect of the relation, rather than a difference in pedigree quality? That is the main “how” in the study of pedigree relations. In order to measure them, controls must be put in place for pedigree quality.

  The thing that is hard for non-eggheads to get past is that the egghead is at least as fascinated by methodology as by the results it yields, but it’s a good thing. Here’s an example. People in our industry talk about opportunity as if it’s reducible to any group of individuals that happen to have something in common. “You have to take opportunity into account,” they say. Most non-eggheads haven’t contemplated what it takes to measure a pedigree relation from the standpoint of its opportunity.

  Don’t worry. As an egghead, I can explain what it takes. Our company’s LyonScore method, which we use in our private consulting, does precisely that, so we have some acquaintance with the issues.

Equalizing the pedigree quality of mares 

  The main issue is that opportunity must be treated in a nominal way. It has to be expressed as a number – of foals, starters, mares, or whatever. In that sense, a unit of opportunity always equals one. Yet, we also know that the real value of a given unit of opportunity – an individual—is determined by its pedigree quality. The effort Bill and Emily apply to measuring the pedigree quality of sires, for which so much data is available, illustrates the enormity of the problem. If pedigree quality varies from one individual to another, how can a certain number of individuals reflect the real opportunity of a group?

  The only solution this egghead knows is to select groups consisting of individuals whose pedigree quality is equal. If it can be reasonably assumed that the members of a group are equal as to pedigree quality, then the number of individuals in the group represents the group’s real opportunity. Some members of the group may deviate to some extent, but the range of deviation needs to be narrow enough to be random to the variable you want to measure. Otherwise, any systematic pattern of inequality among the members of the group threatens to distort the measure of the variable.

 We begin our LyonScore analysis by assuming that the dams of foals by an individual sire of substantial market standing comprise a group that is homogeneous as to pedigree quality. We use a test group consisting of dams that have a given ancestor, and we use a comparative group consisting of all other dams. Then, we compare the proportions of dams that produced superior runners by the subject sire in each group. It’s a simple comparison of proportions showing how well the sire has interacted with any given ancestor. We maintain separate data sets for each sire included in the system.

  This type of analyisis – a simple comparison of proportions – can’t take just any sire as a subject sire. The sire must have substantial market standing based on proven merit. Below that level, too much is left to chance. Furthermore, it won’t do to apply this analysis to sire lines. Combining the mares bred to a sire with the mares bred to his four sons at stud, for example, results in at least five homogeneous groups, probably even more. Those sons all probably contribute superior runners in unequal proportions with the sire and with one another. Most sons contribute foals in large disproportion to their contribution of superior runners. Also, the breeding of those sons will certainly establish biases for and against ancestors of the collective group of mares. It breaks every rule in the egghead book of rules.

Pedigree quality strikes back

  For each sire, our comparisons yield thousands of measures reflecting his responses to individual ancestors of the mares that produced foals by him. With some ancestors he does better than his average, and with other ancestors he does worse. On that basis, a score can be established for each of the mares that produced a foal by the sire. These scores provide a basis for identifying the mares that are best suited to the sire. 

  The advantage of this method is that it yields precise measures of a sire’s pedigree relations to the individual ancestors of his mates, but it involves a trade-off. Keep in mind that the pedigree quality of the sires equalizes the pedigree quality of the respective mares that have been bred to them. Clearly, the pedigree quality of the sires has not been equalized. 

  You begin to get the picture when you see that the Galileo/Danehill cross performs only around 1.6 times its comparative group while less effective crosses for other sires have much higher multiples. The quality of the sire affects the measures of his pedigree relations from two different angles. 

  First, a sire’s pedigree quality will affect his overall percentage of superior runners. The very best sires will have very large percentages in their comparative groups. Consequently, even for the ancestors they favor the most in their mates, their measures of improvement will be relatively small. Weaker sires have smaller percentages of superior runners in their comparative groups, so the measures of improvement for ancestors most favored by them will be much larger.

  Second, a sire’s pedigree quality will affect the range and variety of ancestors he favors. The very best sires tend to favor a wide variety of ancestors, and this causes even more downward pressure on their measures of improvement. By contrast, the preferences of many sires are distributed among fewer ancestors. A sire with a very narrow range of preferences will tend to have larger measures of improvement relating to those ancestors.

  Variation in the pedigree quality of sires strictly prohibits the use of these measures for comparing different sires with one another. A multiple of 1.6 is only run-of-the-mill improvement for mediocre sires, but for Galileo it’s a big multiple. Those measures are so specific to an individual sire that they have no systematic relation to the larger population of sires. It doesn’t take an egghead to see that taking opportunity into account has a cost. A method can’t have its cake and eat it too.

  This simple little comparison of proportions yields a comprehensive profile that can be used to assess an individual sire for an individual mare; but to suppose that the measures controlled by one sire can be compared with those controlled by other sires is nonsense. 

  Those are the rules for taking opportunity into account when measuring pedigree relations. After the what, the why, the when, and the where of things have all been sorted out and sifted through, it’s up to the egghead to figure out the how. The egghead’s unenviable role is to establish the rules. It’s only natural that people would find us irksome.

  Gilded Age economist Thorstein Veblen was right when he observed that the cause-effect structure of industrial production is alien to the business class. However, he betrays his own alienation from the mechanics of cause and effect when he concludes that engineers should be put in charge. Nobody ever gets put in charge. Charge of things is something taken, not given. What Veblen didn’t realize is that we eggheads are happy for others to take charge, so long as they play by our rules. n

 

Search