EQB – Data Jockey

Data Jockey™

Our 40,000+Thoroughbred Databases and Software for Performance Prediction of Unraced Prospects

What data does EQB’s DATAJOCKEY™ collect?

For the past 30 years, EQB has taken several measurements from Thoroughbred racehorses’ cardiovascular systems using specialized equipment for two-dimensional ultrasound echocardiography. We have also digitized data from slow-motion films of thousands of horses. The resulting database, representing over 50,000 young, unraced horses at select quality auctions and farms, is carefully maintained and includes detailed race records (including date of race, racetrack, race surface, race number, distance, race level, time splits, finish position and earnings). Pedigree, conformation, veterinary exam notes and auction information are also stored with each record.

How should one use these several variables all together at once? How does one rank the relative importance of the different variables?

In the past, due to the lack of biostatistical mathematical sophistication in most veterinary school curricula, and to the large computer power required to process big, multi-variable databases into complex polynomials, studies usually tried to create simple ratios of pairs of variables and mathematically fit curves to them for two-dimensional graphs and equations. That is, they compared only two variables at a time for rankings of variables’ relative importance to one another.

But with today’s computers, a much better way is readily available, and it’s called multivariate discriminant regression analysis. For EQB, this has allowed as high as a 48-axes “space” and higher order polynomial functions in describing statistically significant relationships and the reliability and reproducibility of their calculated probabilities in predicting subsequent racing performance.

Interestingly enough, even with these kinds of sophisticated statistical techniques, the relationships to subsequent performance of many traditionally used criteria for horse selection are still weak. Which is part of why it’s so hard to find an elite racehorse. That’s why EQB also uses non-traditional variables. And that’s why traditional horsemen, even those in the Hall of Fame, rarely have over 10% of what they buy race at the highest levels.

By utilizing EQB’s exclusive DATAJOCKEY™ service, clients take advantage of our massive, unique databases and the models developed from them, to assess a horse’s performance probability — even before they ever have a single race!

So how does EQB use multivariate discriminant analysis to contribute to the racehorse selection process?
Just as most people can’t visualize more than a three-dimensional space, they similarly can’t conceptualize how more than three or four variables work together to form patterns. However, statistical programs can identify patterns in as many dimensions as there are variables to make up a pattern.

These are the same analytical procedures used routinely for human disease, medical device and drug studies.

An example of such a statistical procedure is the multivariate discriminant analysis EQB uses to predict horses’ future racing performance based on cardiac measurements.

Multivariate discriminant regression analysis is especially useful for categorizing things into groups, such as high earners vs. low earners, or sprinters vs. routers. In medicine, similar programs are used to determine a patient’s chances for survival with surgery after taking into consideration dozens of test measurements (there the groups, or “categories”, are, for example: live or die, or perhaps, if there are 4 categories: die, live one year, live 2 years, live 3 years+). When a doctor has 20 different test scores describing a patient’s health, it becomes difficult to summarize what they all mean, since it’s often not black and white. It’s certainly difficult to condense all the different kinds of data available at racehorse auctions into clear purchasing decisions.

Can you visualize a 48-dimensional space (i.e., 48-axes graph)? Can you compare a dozen variables simultaneously? EQB’s technology can.

EQB’s exceptionally large and well maintained database, created over 25 years, allow it to be very precise in the creation of statistical models. For example, when EQB analyzes a 17-month-old horse’s cardiovascular system, it compares that horse to hundreds of others (with detailed, known, subsequent racing histories) of the same sex, that were measured at the same age (within 30 days), and size (height and weight, e.g., within 25 pounds).

As an example of how EQB builds a computer model, it may take 1,000 high earners and give the computer all their vital statistics, such as sex, height, weight, age and several cardiac measurements (or data from biomechanical analysis of slow-motion videos of the way of going). Then it will create groups of terrible or modest performers. The computer then creates sophisticated statistical profiles to describe each group.

When an unraced horse’s data is entered into the computer, the computer compares the new data with that of high and low earners it has already seen in its database. The computer assigns the new horse a probability of belonging to one group or the other. For example, it may calculate that a horse has a 30% probability of belonging to the high earner group and a 70% probability of belonging to the low earner group.

If you could visualize these statistics in a multi-dimensional space, the groups would look like planets (try to picture clusters of data points) in space. The models calculate the new horse’s location (the location of its data point) between the clusters for the high and low earners. Complicating matters further, more than two groups can be compared simultaneously (such as low, medium and high earners), and the “space” can be (and in EQB’s case is) composed of many more than three axes.

Multivariate discriminant analysis as a tool makes previous, simpler, statistical systems obsolete.

These same statistical techniques are routinely used in reporting studies on human drugs and medical devices.

Consultant Team

EQB’s scientific publications and its exclusive technologies were developed with the help of the following team of distinguished professional consultants through 30 years of research (list is incomplete):

David Barlow, Ph.D.

Director, Biomechanics Laboratory, University of Delaware
Dr. Barlow’s work includes high-speed filming and biomechanical analysis of
numerous stakes-class racehorses and of the World Champion Olympic Equestrian, Bruce Davidson. He has also done considerable work with the professional baseball team, the Philadelphia Phillies, and with major manufacturers of a variety of types of sports equipment.

G. Frederick Fregin, V.M.D.

Cardiology and Exercise Physiology, Director of Equine Hospital at University of Virginia. Internationally recognized specialist in Cardiology and Exercise Physiology.
Dr. Fregin has pioneered equine cardiology and exercise testing. He has done considerable research on swimming horses, and has studied the effects of various drugs, including “Bute” and “Lasix” for the Pennsylvania Racing Commission and other prominent equine organizations.

William Turner

Trainer, Seattle Slew, Czaravich and other Champions

Professor George W. Pratt, Jr., Ph.D.

Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology and Tufts School of Veterinary Medicine
Dr. Pratt is a tenured professor at MIT who has developed revolutionary tools for evaluating and predicting a horse’s potential as an athlete and its chances of becoming lame or breaking down.

John R.S. Fisher, V.M.D.

Glendarro Farm, PA
An experienced veterinarian and a trainer of many stakes-class Thoroughbred flattrack and steeplechase racehorses.

Maurie D. Pressman, M.D., P.A.

Dr. Pressman is a Clinical Professor of Psychiatry and an experienced professional in Sports Psychology for top-notch athletes.

Douglas Rabin, M.D.

A specialist in Ob/Gyn with experience in medical ultrasonics.

J. Richard Trout, Ph.D.

Professor of Statistics, Rutgers University

Virginia Reef, D.V.M.

Professor of Medicine, Widener Hospital, Director of Large Animal Cardiology and Diagnostic Ultrasonography, University of Pennsylvania, School of Veterinary Medicine.

Participation as a consultant in EQB research does not necessarily imply endorsement of any or all of EQB services.

EQB’s “scientific” horse selection aids were developed by teams of real scientists and validated in prestigious recognized scientific forums.