Are Racehorses Getting Faster pt 2 -- 17 November 2003 -- Jerry Brown

Okay. We established in parts One and One A that it is not possible to compare thoroughbred racehorses from different generations by comparing the raw times of races, because the track surfaces and cushion depths have changed over the years, and there is no accurate way to quantify the differences.

From “The Nature Of Horses”, by Stephen Budiansky, published in 1997:

“One idea that has been little explored is the relationship between track surfaces and speed in horses. Studies of human runners by the biomechanician Thomas McMahon led to the discovery that a track whose natural springiness is ‘tuned’ to match a runner’s stride frequency can markedly improve performance. The ideal surface should ‘give ‘ as the foot first comes in contact with the ground, then begin to rebound halfway through the stance phase, and reach full rebound just as the foot lifts off. That way the track’s springiness works in concert with the leg’s motions; it allows the runner to regain some of the energy he loses to the environment as his foot strikes the ground.

By the same token a surface that is too stiff rebounds too quickly, which merely adds to the shock of impact; a surface that is too mushy rebounds too late to be of any help. A properly tuned track for runners feels unusually springy to those unaccustomed to it, but runners report markedly fewer injuries and find running on such a surface to be unusually comfortable. When McMahon helped to build such a track at Harvard, running speeds increased by about 2 percent.

A similarly tuned track for horses would be about four times stiffer than the track for humans, and in fact would feel unusually firm to human athletes; calculations suggest that a properly tuned equine track would be one firm enough that the hoofprints of a galloping horse would make an impression about half a centimeter deep.”

Well, if horses had all been running over that imaginary track over the last 20 years, we would be able to compare their times and relative abilities. The problem is, we wouldn’t have any horses left—the reason that tracks for thoroughbred horses are different is that no other athlete moves as fast, is as big, lands with as much impact, and does so as often in competition. So tracks are built with the safety of the athletes in mind-- increasingly so over the years, as horses have gotten bigger and stronger, and studies have increased knowledge of the cause and effect relationships.

There have been quite a few studies done on track surfaces, many of which don’t apply directly to the issue at hand. Mall posted one a while back about moisture content affecting energy returned to the horse when he lands and rebounds. As an aside, it included a passing reference that “Several studies have shown that the composition of the track surface alters the dynamic responses of the soil, and that THE COMPACTION OF THE TRACK SURFACE MAY VARY WIDELY OVER DIFFERENT AREAS OF THE SAME TRACK (emphasis added). Like, for one and two turn races? I’ll be trying to get my hands on those studies when I get time, so Len, this is a heads-up— incoming.

I won’t claim to have done a tremendous amount of in-depth research to write this, but what I did come up with focuses on the improvement of the breed itself, as opposed to those things that could improve the performance of an individual athlete, other than passing comments about the possibility that such things exist. In general, the reasons for improvement in the breed itself can be traced to:

1—Selective breeding for specific racing traits. This is not done with too many other species, certainly not with humans, and a generation for thoroughbreds is about half as long as for humans, so the breed could conceivably improve twice as fast.

2—Nutrition has improved. Humans are getting bigger, stronger and faster with each generation (check out the roster of a pro football team from 1980 and compare it to one from now—Bob Lilly would get pushed all over the field), and there is no reason to believe horses have not as well. If they were weighing and measuring horses regularly we might have a way of checking that aspect of this.

The following is from “The Genetics of Thoroughbred Horses”, an article by Patrick Cunningham that appeared in Scientific American, May 1991. Mr. Cunningham was professor of animal genetics at Trinity College in Dublin, head of animal breeding and genetics and deputy director of the Irish National Agricultural Research Institute, and director of the Animal Protection and Health Division of the Food and Agricultural Organization (FAO) of the United Nations.

“To assess the racing ability … we used Timeform ratings, which we consider to be the best available quantification of performance for horses in Britain and Ireland. … Timeform was established in 1948 by the late Phil Bull, a remarkable English mathematician turned punter (or gambler, for American readers). ..

The entire selection and breeding process in Thoroughbreds is founded on the belief that racing performance is inherited. Attempts to analyze the genetics of performance in a systematic way have involved some distinguished names, including Charles Darwin and Francis Galton. It is only in recent decades, however, that good estimates of the heritability of performance based on adequate data have been produced.

Toward that end, my colleagues and I have made several major analyses of Timeform data, the most recent of which included the end-of-year records for 31,263 three year olds that raced between 1961 and 1985. We have attempted to measure whether groups of half brothers or half sisters have ratings that are more alike than those of randomly assembled groups. Similarly, we have also looked at the extent to which the ratings of parents and their offspring resemble one another more than do those of random pairs of individuals selected from two generations.

Our best estimate says that track performance, as measured by the Timeform rating, is about 35% heritable. In other words, about 35% of all the variation that we observe in track performance is controlled by heritable factors and the remaining 65% by other influences, such as training and nutrition. If a mare and stallion are each rated 10% higher than the average for the population, then we can expect that their offspring will have ratings that are on average about 3.5% higher. Bear in mind, however, that there is not a straightforward correlation between a horse’s handicap rating and its actual speed.

With that performance heritability figure in mind, Barry Gaffney of Trinity College and I sought to estimate how much the performance of Thoroughbreds should be improving over time, based on the idea that the horses with the best track records are favored for breeding. The average generation length in Thoroughbreds is about 11 years. Approximately 6% of colts and 53% of fillies are selected for breeding. Putting this information together with the estimated heritability of performance, we calculated that, on average, genetic improvements in Thoroughbreds should raise the mean Timeform ratings by 0.92 unit each year.

We then tried to verify that genetic changes in the Thoroughbred population were taking place at the predicted rate. Working with 11,328 Timeform ratings for three year olds, we estimated the relative genetic merit of the stallions born in the years 1952 to 1977. Our analysis showed that although the average genetic value varied somewhat from year to year (as one would expect), it had a steady upward trend that averaged 0.94 Timeform unit per year. This figure was remarkably close to our prediction and confirmed our belief that selection is steadily improving the average racing performance in the population.”

Cunningham then goes on to wonder why the times of the English classics are not improving as much as with the “shorter” American classics. I would suggest that if the surfaces are in fact the same as they used to be (at least questionable), the answer probably has to do with the slow early paces of the European marathons. As we see here with “pace” races, a slow pace dramatically affects final time.

A couple of other points about the Cunningham piece. First, five Timeform points equals one Thoro-Graph point. At the rate of improvement he discusses, horses would improve a little more than 18 Timeform points over 20 years, or a little less than 4 Thorograph points over the time we have been making figures. Second, racing in Europe (at least in theory) is drug free—no lasix, etc. That certainly figures to keep improvement there closer to just the improvement of the breed itself, as opposed to improvement in performance of individuals that are getting help.

Which brings us to the question of improvements that are not inheritable.

1-- Training methods themselves improve over time as they are studied, especially when there is big money in it.

2-- Sportsmedicine is improving, and improving athletes, especially when there is big money in it. This goes for everything from therapeutic assistance to steroids and all the other additions that have helped Bonds, McGwire, Sosa, and all the others who are carrying 30 or more pounds of muscle than Aaron, Mays, and Mantle were. And, as opposed to humans, horses don’t get a vote about what gets put into their bodies.

3— Specific performance enhancers. Interestingly enough, both Budiansky and Cunningham discuss a ceiling of performance for thoroughbreds based on physiological issues, specifically regarding ATP, oxygen, blood, lactic acid etc., and exhaustion. As it happens, the drugs that have come under discussion over the last few years (EPO, milkshakes etc.), deal directly with eliminating this ceiling.

At a more practical level, we all can think of some trainers whose runners improve quite a bit when they get their hands on them. There are two possibilities—the first is that when they run the horses, all the competing runners are getting magically slower, allowing them to win while not improving. The second is that these trainers are getting improvement out of their runners. If that is true, there are maybe 500 thoroughbreds racing at any given time which have been moved up, meaning a significant percentage of the races are going faster than they otherwise would— performance of the breed is in effect improved.

That aside, there are LEGAL drugs that can either be considered sportsmedicine or performance enhancers, depending on how you look at it, and which were not legal or in widespread use 20 years ago. Most notable of these, of course, is Lasix, but a trainer could probably give you a list of others.

So, how do we quantify the improvement? We can’t use raw times, but is there some way we can extrapolate to estimate how much thoroughbreds have improved? Well, we can look at some other breeds.

The mile record for humans when Thoro-Graph began making figures in 1982 was 3:47:33. It is now 3:43:13, a difference of almost 2% of final time (and that represents a lot less improvement than in the previous 20 years, for whatever reason).

But the closest thing we have to a parallel is standardbred racing, since they are horses, and the economic forces driving that industry are very similar to those in our game. While there have been some changes in racing surfaces, harness tracks are for the most part hard and flat, since the injury dynamics are different, and they almost always run exactly a mile, which makes things easy. Just for starters, in 1980 there were 138 miles trotted in less than 2 minutes—in 2002 there were 5,972. Pacers broke 2 minutes 3,760 times in 1980, 42,598 times last year. Records are kept broken out for individual ages, genders, and gaits, as well:

Trotters 2002 1990 1980
2yo c&g 1:53:2 1:55:3 1:57
2yo f 1:55 1:55 1:56:3
3yo c&g 1:51:3 1:52:1 1:55
3yo f 1:52:1 1:52:4 1:56:3
4+ c&g 1:50:4 1:53 1:54:4
Mare 1:51:4 1:54:4 1:55:2

Pacers 2002 1990 1980
2yo c&g 1:50 1:51:1 1:54
2yo f 1:51:2 1:51:2 1:56:1
3yo c&g 1:48 1:48:2 1:49:1
3yo f 1:49:2 1:51:2 1:53:3
4+ c&g 1:46:1 1:49:2 1:52
Mare 1:48:4 1:50:4 1:52:4

Based on the above, and assuming track speeds have stayed relatively constant (a big assumption), standardbreds have cut their times (improved) by about 3.25% since 1980. Using six furlongs for convenience, since one fifth of a second equals one Thoro-Graph point at that distance, and using a base time of 1:10 because it makes the math easy, the same degree of improvement in thoroughbreds would result in running about two seconds and a fifth faster, or 11 points. Which would be an awful lot.

But ultimately, Patrick Cunningham had exactly the right idea. The best way to compare horses from generation to generation is through using accurate performance figures, since their whole purpose is to compare horses which run on different days, over different tracks. The one caveat is this: you can’t do it with figures that use claiming pars that anchor the data base in place by ASSUMING that the breed does not improve over time. That becomes a self fulfilling prophecy by definition—if you decide the claimers can’t improve (particularly ridiculous given the move-up trainers), the figures for the stake horses can only improve if they get better RELATIVE to the claimers.

So, how much have racehorses in this country improved, using the best available means (Thoro-Graph) as a measure? The answer is, not as much as harness horses. If we get the time we’ll run a study division by division (25k older male claimers, etc.), at least for the data from 1992 on that we have stored electronically. But it’s pretty easy to make a rough estimate by looking at two things:

1—The figure it takes to win big races like the Derby (winners from 1982 on are available on this site) is about 5 points faster than it was when we started making figures.

2— Len Friedman posted on the Ragozin site a while back that the reason their figures could be used to compare horses from different generations is BECAUSE they anchor the figures to claiming pars—he had it exactly backward, but it sets up a useful scale. When we started making figures they ran about 3 points slower than Ragozin’s (zero points are arbitrary—you could put it anywhere). While there is tremendous variance race to race, track to track (depending on who is making the “Ragozin” figures), and distance to distance (don’t get me started), our figures now run about 3 points faster on average. So on that basis you could conclude horses have gotten about 6 points faster.

Which is interesting, if you keep in mind Cunningham’s Timeform study, which would indicate that genetics alone figured to improve the breed by almost 4 Thoro-Graph points over this time period. Factor in Lasix, improved nutrition, and everything else, and it certainly seems 5-6 points of improvement since 1982 is about right. Five points at 6 furlongs is one second, which using our 1:10 example means racehorses have improved about 1.5 percent.