On Breeding Values for Thoroughbreds

[brogers]

I sent the following letter to the Thoroughbred Daily News in response to a column by noted industry observer Bill Oppenheim on stallion selection in mating decisions (it was published). It outlines the reason why the industry at large needs scientifically valid breeding values for true genetic improvement to occur.

To the Editor

In Bill Oppenheim column Near Miss, 6/13/12 he raises a fundamental issue that for far too long has been ignored by the thoroughbred breeding industry.

Since the time when the Sport of Kings was literally, the sport that kings participated in, thoroughbred breeding has been a non-random event, based on the belief that there is something to gain from the selection of mates that could not be obtained by random selection of mares and stallions. Seeking an improvement in the process of selection and breeding in thoroughbreds is founded on the belief that racing performance is inherited.

Heritability is a measure of the degree (0 to 100%) to which offspring resemble their parents for a specific trait, which broadly in our case is racing performance. Despite the intensity of selection carried out in the Thoroughbred, race times have not improved any further for at least the past 60 years (Hamori and Halasz, 1959; Cunningham, 1975). Research has however shown that many important traits in the horse are moderate to highly heritable (i.e can be passed on from parents to offspring) indicating that genetic progress from selection is possible if the methods of selection themselves are based on sound principles of inheritance.

My experience has been that in European countries, breeders and their agents place very little importance on the direct speed achievements like speed figures when selecting for breeding, rather their performance relative to their peers of the year and other generations before them with the use of Timeform figures. It seems their belief in subjective measures such as Timeform is well placed as Timeform has a relatively high known heritability of race performance at 35% (Cunningham, 2002). Thank you Phil Bull.

Somewhat conversely, North American breeders use two methods that are linked to that which is known to have low heritability for selection. Earnings, which statistics like the A.E.I are based on, are a very poor measure of racing performance as they have low heritability (~10%) and are not optimally distributed in that a horse running 7th in a Grade One gets no prize money while a horse running 3rd in a claimer at Finger Lakes does even though the horse running 7th is most likely a significantly better horse.

More importantly race time has also been used as a measure of performance in Thoroughbred racehorses but the heritability of race time has been reported as low as 9% to 16%, depending on the distance of the race (Moritsu et al., 1994, Oki et al 1995). This leads us to the crux of the problem, in the U.S in particular, that being the erroneous use of race time and race time based handicapping figures for selection of both breeding stock and matings. Noting above that race times as a whole are generally a lowly heritable measure, and with a caveat that the heritability of any time based speed figure is not known (itself a problem!), the use of a handicapping figure derived from race times such as Beyer numbers for breeding selection does not appear to be the most suitable method of genetic selection and could well partially explain why improvement in the North American thoroughbred has not been made and to a certain extent why European breeders, relying on the more heritable Timeform figure, have continued to breed elite racehorses despite their breeding stock being significantly depleted at times of economic downturn.

As breeders we keep hearing year after year how each crop is getting slower in terms of the feature races such as the Kentucky Derby, Preakness and Belmont being won with lower and lower speed figures, yet we don’t think for a second that selecting our stallions and mares on speed figure numbers looks to be the cause! At its core it must be understood by breeders that the fact that a stallion ran a “1 on the sheets” or a “124 Beyer” should mean nothing to them when judging him as a stallion, or a mare as a potential broodmare for that matter, as the heritability of this figure is unknown but because it is based on time, most likely to be very low. Breeders certainly aren’t getting what they think they are paying for when they select a stallion or mare based on a speed figure. This leads us to what animal geneticist Dr. Betrand Langlois stated so clearly, “Why should one pay more for this or that pedigree or high performing sire if the performance itself is not heritable?”

Despite the fact that breeding and racing of thoroughbred horses has been a well-organized industry for well over a century now, it is a poor reflection on the industry at large that it has not yet come to an understanding of the best measure of racing performance or ability and the degree of inheritance of this racing ability. We use figures for selection that are built for handicapping, not breeding and because of this we are actually selecting on a measure with low heritability and act surprise when the breed doesn’t get an faster or sounder! Breeders worldwide are effectively trying to sculpt a masterpiece by using a chainsaw, all the time not even knowing what a masterpiece actually looks like.

Creating a universal breeding value/figure as opposed to using a handicapping figure for selection is not a difficult task, nor is it expensive and it should be very easy to disseminate to breeders worldwide. We have the data (we actually have one of the most data rich breeds in the world, why we don’t use this data more constructively is a mystery) and the methods have been thoroughly proven in other livestock and plant industries to great effect. Much, much smaller breeds such as The Danish Warmblood, Irish Sport Horse, Dutch Warmblood, French Sport Horse, Swedish Warmblood, Holstein, Hanoverian, Icelandic Toelter, German Trotter and Swedish Trotter Associations all use breeding values and have shown significant increase in genetic progress since their creation and implementation.

Thoroughbred registries around the world seem to have a similar mission statement – “dedication to the improvement of Thoroughbred breeding and racing.” While something like this will never make a Round Table for discussion or an expensive consultants report, and thus not get the attention it desperately deserves, actually having a scientifically sound framework of genetic merit is the essential starting point for any improvement of the thoroughbred to happen in the first place.

On the first Saturday in May and indeed each year on Belmont Stakes day for the “Test of Champions”, we can decry as much as we like about how we are not breeding horses to get a route of ground but only if we actually scientifically measure and disseminate these traits, and that could include not only speed over distances but also other traits like fertility and soundness, can we actually start to make progress. To get anywhere, you have to know where you are starting from and right now, by using speed figures for selection as a de-facto of a truly heritable measure, breeders have no idea where they are, where they are headed, and are probably doing the opposite of “improving the breed”.

Byron Rogers
Performance Genetics LLC
Pedigree Consultants LLC

[Tappiano]

Nicely written but isn't dosage supposed to be a predictor of stamina and performance? Granted it does not predict whether it can enable an improvement in time but it's still shown on pedigrees so it should not be discounted.

The Japanese are also doing a good job at identifying the right names in pedigrees to include in their program as do the South Koreans. They, like the Europeans are trying to improve the breed. Improvement does not necessarily equate to faster but it sure equates to stamina. Perhaps improving the stamina will allow for an increase in speed because the horse will be more efficient? Maybe it's as simple as breeding like to like because then you know what you get physically.

That brings us back to the breeders in this country who as a whole just don't care. How many horses do you see in the November sales catalog and go WOW that's a great pedigree and she's in foal to the right stallion? More likely it's "why is that unraced mare out of an unraced mare out of yet another unraced mare who happens to be a full sibling to a G1 stakes winner bred?"

There will never be a change in what breeders do as long as they can still sell their weanlings and yearlings profitably.

[brogers]

Nicely written but isn't dosage supposed to be a predictor of stamina and performance? Granted it does not predict whether it can enable an improvement in time but it's still shown on pedigrees so it should not be discounted.

Dosage is in no way related to breeding values.

As an example, Let's say that you have a broodmare that is by Rock Slide out of an unraced mare who is by Secret Claim out of mare X. This horse was useless on the racetrack.

And then you have another broodmare by Mineshaft out of a stakes winning Gone West mare who is out of the same mare X. This horse was a G1 winner.

On dosage, these two have the same scores as they have the exact same elements in their pedigree with Rock Slide being a full brother in blood to Mineshaft and Secret Claim a full brother to Gone West and them sharing the same second dam. Effectively they are full relations in blood.

If you had a breeding value, they would have completely different values however when they were unproven yearlings and then again as broodmare prospects. The breeding value would consider the race performance of the two horses as well as the race (distance and class) and production (distance and class) performance of their immediate relatives (sire, dam, grandsires and grandams). Thus they would have two wildly different breeding values and would actually, rightly, place the genetic merit of the Mineshaft mare well above the Rock Slide mare.

[Tappiano]

Dosage is one tool in determining whether a horse is likely to get a distance of ground but it's no more or less useful than anything that's out there already. You reference a problem with stamina and I mentioned dosage.

Your example is something that everyone already has access to but again, how is it going to make any difference to a marketplace that demands a physical that resembles speed and demands a ready made two year old, not one's that should be turned back out into a pasture to grow.

The marketplace is "quick" to hop on the bandwagon when it comes to a "nick", ie, El Prado over Forty Niner (Rachel Alexandra) but not every single stallion with El Prado and every dam by Forty Niner is a good match physically. In your letter, would it be better to go with the one who has the higher value who does not match physically or the lower one who does?

I love a good discussion on this subject because I'm all for improving the breed but perhaps the starting point is managing the expectations of buyers?

[Joltman]

thanks Byron

Article raises some really good questions. I think the assumptions that certain factors contribute to a 'quality racehorse' would seem to be sound - i.e. that a horse can not receive something its sire/dam cannot pass on. The problem is that there are SO many variables.

But the general thinking seems to be that:

1. Speed (and speed figures which reflect speed over a distance in real race conditions accounting for track, etc.) impliesthat generally

the fastest horse wins,
faster horses run in higher classes
higher class runners earn more money
those runners can pass on that ability

So, stamina is generally ignored, but

2. Stamina - lots of horses can go 21 and change. How long can they go 23 or 24 a lot farther is probably much more important. Tappiano's remarks on stamina get the star.

3. Soundness - A potential sire's potential aptitude is almost worthless if a horse cannot meet a minimum number of high-class (stakes-graded) starts to get on the radar screen before the wheels fall off. A horse may have the raw aptitude to run a mile and a quarter in 1:58 and never have a chance to show that. The soundness issue will be influenced by everything from the (truly) freak accident, conformation weaknesses, dietary practices, vet practices, early retirement to stud, and, oh yes, training practices. The top trainers may preferentially select high-priced horses who will never remain sound and racing for any of the above reasons, yet these are the ones most likely to contribute to the gene pool.

4. Conformation - weaknesses obviously influence, but also extraordinary strengths perhaps like the large heart. This will enable the quality horse to race faster/longer leading it up the ladder in class and value. The bio-mechanics guys probably have an edge here where it's not just look, but measurable. As I have opined elsewhere, I believe that sales buyers (since Lucas) have redefined what a 'good looking horse' is, and many quality runners from decades past would be panned if they showed up in a sale today. (too small, fine-boned legs etc.) Take a look at the Classic Sires in the Ttimes some time and see their 'faults' jump out.

5. Misc. - things which are emotional like 'fashionability and hype' that led people to throw $100k first season unproven Smarty Jones, will skew the gene pool.

I would disagree about earnings indices if, as some are interested in this as a business, that earnings are essential(!) And some sires (and BM sires) tend to improve whatever they are bred to. And there is a general correlation of running index to stallion success (D Dink wrote on it back in the 90s). The analysis of the G1 runner running 7th vs the claimer is too simplistic because yes, in one race that is true, but in terms of class, the next race the Graded runner might drop and win against G2 company and his earnings will quickly eclipse the FL claimer over time. That is not to denigrate the hard-knocking claimer that has something going for it, namely soundness. In this game of tortoise and hare, the hare wins, but a good tortoise is Ok to have as well. A bad hare or a bad tortoise is what needs to be avoided.

so those are a few thoughts. A breeding index, in my opinion, would integrate and weight many different factors, but if enough were included,could lead to something statistically predictive. The nicks stuff you guys do is (IMHO) one such factor and of some value in context.

jm

[jellac]

Interesting letter Byron - not being aware of the degree of a performance trait's general inheritability when contemplating a mating throws into question to what degree one is actualy practicing 'selective' breeding, no matter how much homework one puts into reading all the stats, etc.

One aspect of this that I think is worthy of considering is how 'productive' the female family is under a prospective mating. I'm not talking about sheer numbers of foals or even sheer number of foals that race and/or win...but rather the overall quality of that female line, tail female in producing quality runners/winners. Further I'm talking about a female family's proven ability to sustain that level of productivity for several breeding generations. I also understand taking a risk on an individual that arises from a less than 'productive' appearing female family when that individual is outstanding in her own performance record on the track and there is evidence of the sporadic production of quality individuals in an otherwise sparse female family. I'm not aware of any method of 'measureing' this female family's robustness in the breeding shed. It seems rather to be a qualitative art in the review of pedigree for prospective mating and much in the eye of the beholder (or mind of the reviewer). Basically I'm interested in any measures that exist by which one can compare how frequently a given female family behind an as yet untested/unproven broodmare has been/can be? Are you aware of any such measure that allows one to compare the productiveness of one female family over another in a quantitative manner?

[brogers]

thanks Byron

Article raises some really good questions. I think the assumptions that certain factors contribute to a 'quality racehorse' would seem to be sound - i.e. that a horse can not receive something its sire/dam cannot pass on. The problem is that there are SO many variables.

But the general thinking seems to be that:

1. Speed (and speed figures which reflect speed over a distance in real race conditions accounting for track, etc.) impliesthat generally

the fastest horse wins,
faster horses run in higher classes
higher class runners earn more money
those runners can pass on that ability

So, stamina is generally ignored, but

2. Stamina - lots of horses can go 21 and change. How long can they go 23 or 24 a lot farther is probably much more important. Tappiano's remarks on stamina get the star.

I think that in order to have a universal breeding figure you would need to break down that figure into distance categories like say 5-7f, 8-10f and 11f+ and have values for each of these categories. Some racing environments select for different things so having one figure is not going to help a lot but having three distance specific figures would. This would also highlight the deficiencies of racing populations in developing stamina which you and Tappiano rightly point out, looks to be an issue in the North American breed.

3. Soundness - A potential sire's potential aptitude is almost worthless if a horse cannot meet a minimum number of high-class (stakes-graded) starts to get on the radar screen before the wheels fall off. A horse may have the raw aptitude to run a mile and a quarter in 1:58 and never have a chance to show that. The soundness issue will be influenced by everything from the (truly) freak accident, conformation weaknesses, dietary practices, vet practices, early retirement to stud, and, oh yes, training practices. The top trainers may preferentially select high-priced horses who will never remain sound and racing for any of the above reasons, yet these are the ones most likely to contribute to the gene pool.

You could create a value for soundness which would probably be the most important value created. Remember we are talking about a population figure here so the freak accidents would tend to be normalized within this figure.

I find it interesting that on this forum and others breeders and owners display a high desire to increase soundness in the breed, but nobody knows how to scientifically measure it, or how to properly breed towards it (I'm ignoring the rants of Louis which have ZERO scientific basis)

4. Conformation - weaknesses obviously influence, but also extraordinary strengths perhaps like the large heart. This will enable the quality horse to race faster/longer leading it up the ladder in class and value. The bio-mechanics guys probably have an edge here where it's not just look, but measurable. As I have opined elsewhere, I believe that sales buyers (since Lucas) have redefined what a 'good looking horse' is, and many quality runners from decades past would be panned if they showed up in a sale today. (too small, fine-boned legs etc.) Take a look at the Classic Sires in the Ttimes some time and see their 'faults' jump out.

This isn't something that a value could measure unless everyone had biomechanical measurements as a requirement of racing and it was measured in a standardized way.

I agree with your assessment of how the type has changed here in North America. The sprinter is asked to go too far for what it is. If you look at the horses that line up in the Kentucky Derby and then go and look at the horses that are in the English, Irish and French Derby and they are completely different types. Sure they are slightly longer races, but their "classic" type even for shorter races is completely different.

[brogers]

One aspect of this that I think is worthy of considering is how 'productive' the female family is under a prospective mating. I'm not talking about sheer numbers of foals or even sheer number of foals that race and/or win...but rather the overall quality of that female line, tail female in producing quality runners/winners. Further I'm talking about a female family's proven ability to sustain that level of productivity for several breeding generations. I also understand taking a risk on an individual that arises from a less than 'productive' appearing female family when that individual is outstanding in her own performance record on the track and there is evidence of the sporadic production of quality individuals in an otherwise sparse female family. I'm not aware of any method of 'measureing' this female family's robustness in the breeding shed. It seems rather to be a qualitative art in the review of pedigree for prospective mating and much in the eye of the beholder (or mind of the reviewer). Basically I'm interested in any measures that exist by which one can compare how frequently a given female family behind an as yet untested/unproven broodmare has been/can be? Are you aware of any such measure that allows one to compare the productiveness of one female family over another in a quantitative manner?

The closest I have seen to what you are asking for was developed by the Broadbent's of BRIS. It is a female family index (FFI). You can get more information on it here.

http://www.brisnet.com/library/indices.pdf

Now this comes with a caveat. The BRIS database is far from complete in terms of international racing so if your mare has any foals in her first three dams that have international performance then the FFI isn't going to be that good a reflection of the actual merit of the female family.

[Joltman]

3. Soundness - A potential sire's potential aptitude is almost worthless if a horse cannot meet a minimum number of high-class (stakes-graded) starts to get on the radar screen before the wheels fall off. A horse may have the raw aptitude to run a mile and a quarter in 1:58 and never have a chance to show that. The soundness issue will be influenced by everything from the (truly) freak accident, conformation weaknesses, dietary practices, vet practices, early retirement to stud, and, oh yes, training practices. The top trainers may preferentially select high-priced horses who will never remain sound and racing for any of the above reasons, yet these are the ones most likely to contribute to the gene pool.

You could create a value for soundness which would probably be the most important value created. Remember we are talking about a population figure here so the freak accidents would tend to be normalized within this figure.

I find it interesting that on this forum and others breeders and owners display a high desire to increase soundness in the breed, but nobody knows how to scientifically measure it, or how to properly breed towards it (I'm ignoring the rants of Louis which have ZERO scientific basis)

.

I think a soundness metric would not be difficult, just a raw number of starts would be a beginning. It would also be desirable to have a class component though. There are a number of Class ratings out there (say the one Brisnet uses). Then count the number of starts a horse has at different class ranges (say comparable to the Graded, stakes, allowance, high claiming, low claiming). That way if a horse runs 30 times and 2 are at (legit) allowance levels and the rest at low claimer it's not the same as the horse consistently running at stakes level or higher for say 20 starts. Those at higher levels are often not run as frequently for a variety of reasons so the metric might compare to average # of starts per year (or by foal crop), (at that level) for the total pool of runners in that class. Since you guys have access to the Jockey Club database you could probably play with this and have a lot of fun with it.

jm

[Sylvie Hebert]

How will "speed" be calculated if as often seen low level claimers often win in times faster than the feature races on the same program? And then we get into suface variations caused by weather,wind,ability of maintenance crew,riders,shoeing,weight...and as stated before,one occurence of record breaking speed does not mean much(How about these sales times?)Wow very difficult and tricky....

[Barcaldine]

It's a complete waste of time to assign absolute values to racehorse performance. They aren't machines whose outputs are measurable and consistent. They are animals which perform at different levels (speed, surface and distance) every time they step on a racetrack.

The problem these assigned values create is a false security in their reliability. Dosage and Nicking are the two most blatant attempts in our industry to claim statistical validation for their unscientific commercial products.

[griff]

Byron

Did the print your letter?

griff

[wet sail]

Byron
Breeding values have made huge increases in productivity in sheep cattle and pig breeds.
If these values were to work worldwide would you not have to make allowances for the US racing using drugs that are banned everywhere else ie lasix .As this is not a fair comparison .

[griff]

I wrote a letter to the Daily Press editor a couple of weeks ago advising people that Obama is a Manchurian Candiate and it was published

griff

[ides of ice]

What about the fact that the overall gene pool for TB's has gotten smaller. ND and Mr.P are in almost every pedigree. The diversity from decades ago is almost gone.