Matchem Genetics

Understanding Matchem Genetics


We are going to talk about what Matchem Genetics does and how it does it and delve into the science behind it, we hope that you enjoy the series and if you have any questions please let us know and we will try to address them if we can. Before we start talking about cardiovascular scans and genetic markers and the like, it is worth first discussing the problem that we are trying to tackle so that we can better understand why we develop the prediction models that we use to select racehorses and how this puts you as an owner in a better position to realize racetrack success.

The Thoroughbred industry is a ‘rare event’ industry, which makes prediction of future outcomes difficult, but not impossible. Somewhat similar to fraud detection in credit cards, 95% of the yearlings we look at will be limited in their athletic ability, while only 5% will be stakes quality standard. Compounding this, or making it a little harder to predict outcomes, is that ‘trainer effect’ the combined effect of the trainer, their track work riders and associated veterinarians and form analysts who help get the horse to the race to win, has a large influence on the outcome and ultimately a horse only has to beat who turns up on the day to compete. It is why Champion Thoroughbreds only selects the best trainers to take on their horses, and places them where they think they can excel, but why at Matchem Genetics we see some horses that we don’t think are that good by the measurements we take, turn out to be good performers
On that idea, when we go to a yearling sale there are four possible outcomes from the yearlings we measure at the sale:

• True Positives – Horses that we predict to be fast, and they subsequently turn out to be fast.
• False Positives – Horses that we predict to be fast, but they turn out to be slow.
• True Negatives – Horses that we predict to be slow, that are indeed slow.
• False Negatives – Horses that we predict to be slow, that turn out to be fast.

Obviously when you consider the above we are trying to minimize false positives and false negatives. Most importantly, because we know that only 5% of the horses at a sale will end up as elite horses, we want to reduce false negatives and not miss a good horse. The trade off for this is often an increase in false positives and a loss of precision, or strike rate as we call it.
When it comes to strike rate, or precision, one other factor must be noted. If we went to all major yearling sales (Easter, Karaka, Magic Millions, Melbourne Premier) on average we are going to see 5% of those horses turn into stakes winners or high level runners. Thus, if a trainer or a syndicator goes to a sale and buys 20 horses, by random chance one of them must be an elite level horse. When you see a syndicator who is buying at these sales and can’t even get one stakes winner from 20 selections it tells you they are doing something dynamically wrong as a dart thrown at a board would do a better job!

The other aspect in building prediction models that we must consider is Recall. This metric describes the percentage of horses we select as potentially elite, when compared to all elite horses offered in a sale. As an example, if we are at the Magic Millions there will be some 750 yearlings catalogued and we know that there will be 40 elite ones in that 1000. If our prediction model selects 50 of the 750 (7%) that it likes as potentially elite horses, and in that 50 it turns out that there were 10 elite horses then our Precision or or strike rate is 20%(10/50), but our Recall is 25% (10/40), that is we only positively selected 25% of all possible elite runners at the sale. We work hard to build models that balance Precision and Recall.

The last wrinkle that we need to consider is the percentage of horses that the model likes, when compared to the percentage of all horses offered at a sale. From the example above, you saw that this was 7% (50/750). Generally speaking we try to build our models so that it is positively selecting for between 5-10% of the catalogue (this varies due to overall sale strength) so as to give Jason and Suman a chance to buy the volume of horses that they bring to you, their clients, each year. The bar that we set is fairly high and we build models that have a projected strike rate of 20% which means that one in five of the horses they buy should end up a stakes winners, but it also means that we have to work really hard and test a lot of horses to bring the right horse into the Champions fold.

Matchem Heart Scores


Cardiovascular parameters measured by echocardiography have been of interest to exercise physiologists due to the relationship between the capacity of the cardio to deliver oxygenated blood to the muscle. The theory to prove was that a strong relationship exists between cardiac dimensions and peak oxygen uptake (V02Max) in thoroughbreds with the latter being a measure of performance (that is horses with high peak Vo2Max scores are as a group better racehorses with higher earnings than those with lower peak Vo2 Max scores).

There are two different ways to measure an equine heart– ECG and M-Mode ultrasound. The ECG method was made popular in the 1970’s by Steel, et al with their paper “The Inheritance of Heart Score in Racehorses” onlinelibrary.wiley.com/doi/10.1111/j.1751-0813.1977.tb00237.x/abstract? and Dr Stewart with his paper “Heart Score Theory in the Racehorse” onlinelibrary.wiley.com/doi/10.1111/j.1751-0813.1981.tb00551.x/abstract While immediately popular, and used by some leading trainers in Australia and New Zealand for some time, follow up studies by Leadon, et al onlinelibrary.wiley.com/doi/10.1111/j.2042-3306.1982.tb02347.x/abstract and more recently Lightowler, et al /www.researchgate.net/profile/Elisabetta_Giudice/publication/8349952_Echocardiography_and_electrocardiography_as_means_to_evaluate_potential_performance_in_horses/links/55351bd70cf216f6b443d7fb.pdf failed to show the link that Stewart and Steel had shown with performance and in fact Lightowler proved that ECG was generally a bad measurement of cardiac output.

Unlike ECG’s, M-mode echocardiography of the left ventricle of equine hearts has proven a relationship between their measurement and racetrack performance. In addition to Lightowler, et al above, Seder, et al proved a relationship between the measurements taken as a yearling and two year old and racetrack performance in Thoroughbreds www.j-evs.com/article/S0737-0806(03)70068-9/abstract , Young, et al found relationship between the left ventricle of thoroughbred racehorses and performance jap.physiology.org/content/99/4/1278 and maximal oxygen uptake onlinelibrary.wiley.com/doi/10.1111/j.2042-3306.2002.tb05467.x/abstract?, while more recently Gur and Matur found a link between cardio parameters measured by M-mode and the velocity of a horse at maximal heart rate www.j-evs.com/article/S0737-0806(12)00158-X/abstract .

So what does a good cardio look like?

Above is the m-mode cardio scan of Champion Thoroughbreds very own Lightinthenite, taken a few years ago. Looking at the image, you can see where the heart contracts, and where it opens. At the point that it is completely open Lightinthenite has an internal measurement of 13.1cm. He was a multiple stakes winner and earned over $625,000 and would be our ideal of a top class miler cardio.


Now let’s take a look at this image of another horse of the same age at the time we measured Lightinthenite. This horse was a little bit nervous when we took the scan so his heart beats are a little closer together, but you will see that in the same internal measurement he is just 8.98cm. This horse won 2 of his 10 starts and just over $30,000 in earnings.

To properly understand the 4cm difference between the two you need to understand that the angle of the beam from the ultrasound ‘slices’ the cardio in half, like taking a knife to an orange, and we are just looking at the measured difference at that point, however the difference between the two is throughout the cardio, top to bottom. This difference sees Lightinthenite have a Left Ventricular Mass (g) calculation of 5550 and a Cardiac Output (Cardiac output is the volume of blood pumped by the heart per minute (mL blood/min)) of 7923 while the other horse has a LVMass of 4210 g and Cardiac Output of 4565 ml/min.

You can probably guess from the above that the larger heart that is pumping a lot more blood in every heart beat is better, but it is important to note that it is, but only relative to the size of the horse. A large heart in a small/light horse isn’t ideal as the cardio is a muscle that gets fit and in order to get a large heart fit you need to work the horse hard, which often breaks down the smaller/light horse. Equally, a really smaller heart, in a big horse just about guarantees you have a slow horse!

Next week we will talk about why we take biometric measurements and how we relate the measurements to other horses in the database, and why ‘heart girth’ doesn’t actually relate to heart size but another vital aspect to equine performance.

Important note Regarding Matchem Genetics and Cardiac Reports on Racehorses:

The purpose of Matchem Genetics cardiovascular scan is to relate the cardiovascular measurements taken to athletic performance, not as a diagnosis of a disease state. No representative or employee of Matchem Genetics is authorized to give veterinary opinions of any kind or to make any representation whatsoever about the health or disease state of any horse based on its cardiovascular scan, unless they are themselves a licensed veterinarian.