Of Math and Men (and Horses), part II

Click here to read part I

Things get really tricky for a system when lack of opportunity forces a default to deeper generations for analysis (which happens less frequently with TrueNicks than with hypothetical systems that depend on a much smaller-and problematically incomplete-sample size).

It's one thing, for instance, to rate Storm Cat when mated to mares by Mr. Prospector, but quite another to rate an untried son of Storm Cat with a mare by a modest grandson of Mr. Prospector. This is because of a couple of factors.

First, the more successful a stallion or broodmare sire is, the more representation he will have in the population, and those representatives will vary greatly in quality and class.

Second, the more successful a cross becomes, the more it will be copied, and by the nature of things, it is going to tend to be copied more frequently using inferior rather than superior stock.

As a result, the strike rate of our example cross will be dragged down by the number of times that it has been tried using uncommercial sons of Storm Cat with mares by modest grandsons of Mr. Prospector.

There is a way around this, however. If we consider a group of mares bred to a stallion, we find that they have generally been bred to other stallions of similar quality. Expensive sons of Storm Cat will generally have been bred to high-quality mares, with less expensive sons of that horse bred to mares of lower quality.

The mares bred to the expensive sons will generally have been bred to other leading stallions, and the mares bred to the less expensive sons, to other stallions at their own level.

Therefore, the group consisting of "mares bred to Storm Cat and his sons" is going to have been bred proportionally to a range of other stallions at a similar level. The reverse of that equation would apply to the group "mares by Mr. Prospector and his sons and grandsons which have been bred to Storm Cat and his sons" with regard to their other mates.

So, when we assess how well Storm Cat and his sons have done with all mares by Mr. Prospector and his sons and grandsons, compared to how well those mares have done with all other stallions, some valid statistics emerge.

The development of TrueNicks provided plenty of opportunity to test the validity of the premise. With the input of The Jockey Club, we were able to evaluate and calibrate the rating on a test group of over 100,000 horses, representing two entire recent U.S. foal crops.

This produced some very significant results. For example, more than a third of all stakes winners earned a rating of A++, and two-thirds of stakes winners are rated A or better. Against this, less than one third of the whole population is rated at that level. At the other end of the scale, horses rated D or F make up more than a third of the population (actually 36%) , but supply only 8% of the stakes winners.

What is even more interesting is that the system performs well when studying horses bred in numerous different geographic locations, or even from different eras going back as far as the immediate post-World War II period. (Watch out for a coming article: The Test of History - Grading the Greats.)

Filed under: , ,

comments powered by Disqus