The Doctor Continues

In response to some great comments and questions on the Upper Limits Frank makes a good point about normalizing. If handicappers tend to keep the average TIMEFORM rating the same across years, then the average TIMEFORM rating of the entire population cannot improve, by definition. Another example of this is IQ score. The average IQ score is, by definition, equal to 100. So, the fact that American children’s IQ scores have stagnated should not be a signal that we need to overhaul the education system and neither should the fact that half of all Americans earn less than the median income indicate that the economy has tanked (though both conclusions may be true). We can easily mislead ourselves with our measures and our statistics if we are not careful. So, the question is: can we make statements about performance trends in the thoroughbred population despite the normalization of scores? Can we predict increases in racing performance based on data that are normalized? Despite normalization across generations, it is still possible to measure the covariance of relatives. If there is a genetic component to performance, then the best horses will have the best offspring and the worst will have the worst (a positive covariance). This is all we need to measure heritability. (Adding or subtracting to the TIMEOFRM scores to normalize will not affect any measures of variance or covariance and multiplying by a scalar to normalize the TIMEFORM scores will cancel out when measuring the heritability. It wouldn’t even matter if we used one panel of judges for the parental generation who tended to inflate TIMEFORM scores and another panel for the offspring who tended to give lower scores – so long as they keep the rank order the same and a similar distribution of scores). Then, to measure the selection differential, one can compare the TIMEFORM ratings of the selected parents with that of the horses who are not selected to breed. The product of these two will give you the estimated TIMEFORM increase in units of pounds-handicap. This predicted TIMEFORM score increase may not actually be attained because of normalization, but the increase in performance should be detectable. Similarly, cavemen IQs were, on average, 100 – but we would still predict observable intelligence increases despite the normalization of IQ scores across generations. Another point Frank makes is to compare thoroughbreds to pitchers. A pitcher’s elbow and shoulder are right at the threshold of failing every time he throws a fastball. The pressure generated is just nearly enough to shred the tendons and the ligaments. So, while muscles can be trained (and even perhaps bred) to improve, it’s the tendons and ligaments - which don’t respond in a similar fashion to training - that set the cap on pitching performance (at least for fastballs). So maybe tendons and ligaments can’t be trained and can’t be bred for improvement. I know that horses store a lot of elastic energy in the tendons of their legs for running but there are still many other factors that go into distance running besides these and it’s these other factors that could respond to training and selection. One thing that has surprised me about some of the comments is that they seem to be saying that the training has gotten worse because the money is in breeding – not in racing. But isn’t there money to be made from breeding to a record-setting fast horse?