Using anecdotal evidence, many racehorse trainers will tell you that bigger horses might tend to be less sound horses but they tend to be better horses. While this may be a function of the worldwide phenomenon of compression of weights by handicappers, we had noticed that one of the variables that our model uses in determining racetrack potential is the estimated body weight of the horse, or more specifically, the standard score generated by the estimated body weight when compared to horses of similar age and same sex.
At the sales and in the field we use the weight estimation algorithm proposed in the Equine Veterinary Journal by Staniar, et al (2004, Weight prediction from linear measures of growing Thoroughbreds) which we have found to be significantly more predictive of the actual weight of the horse when compared to other popular weight estimations. From there we generate a Standard Score for the horse when it is compared to all other horses of the same sex, and born within 30 days either side of the subject horse at the time of measurement. What the standard score does is tell us, in terms of how many standard deviations a horse's weight is above or below the average weight of horses of the same sex and age. All things being equal in terms of age and sex, the larger and heavier horses tend to be better racehorses.
Obviously before we got to that conclusion, we had to have evidence that estimated weight was indeed related to racing class, and if so, by how much?
Above is a chart that shows 106 colts of the same age. Across the bottom (X Axis) you have their estimated weight using the Stanair algorithm. The Y Axis has a measurement of racing class with horses that score a 1 being unable to break their maiden and horses scoring a 10 being Group/Graded Stakes winners. The ones in between are ranked relative to the top and bottom. It is a little bit of an abnormal distribution of performance in that you would normally see more horses scoring 4 to 6 in real life, but adding those horses makes little difference to the relationship.
The Pearson (r) correlation for the relationship is 0.53. Pearson's r is a sample correlation and measures the degree to which one variable is correlated to another in the sample population where 1 is total positive correlation, 0 is no correlation, and −1 is total negative correlation. A score of 0.53 is in terms of what we are looking at, indicates moderate correlation between estimated weight and racing class. So the trainers that suspect larger and heavier horses tend to be better ones are correct but the size of the horse is relative to horses of the same sex and age.