The following was originally
published as three separate articles in The T I A Newsletter issues
of November, 1994, through February, 1995.
Part One: Basic Genetics
Nicking, once considered
a somewhat interesting theory worth no more than an occasional mention,
has come to dominate Thoroughbred breeding to an extent dangerously out
of proportion to whatever value it may offer.
Not long ago the owner of a stakes-winning
mare retiring from the track told us he called a respected Kentucky bloodstock
agent for help in deciding on a mate for her and the man said he couldn't
make a recommendation until he could find out what score the nicking experts
would give the prospective foal.
Soon after that, the manager of a major California
farm told us that breeders calling for bookings to the most popular of
their stallions don't ask about the horse's conformation, size, or temperament;
they just ask about the nick.
Breeders everywhere beg to be told which stallions
provide the magic nick for their mares and pay unregulated gurus to turn
the key that unlocks the door to greatness. Purchasers of racing prospects
increasingly rely upon nicking scores provided--often at substantial fees--by
the high priests of the new religion.
Those nicking experts have become so powerful
that, as noted above, even knowledgeable horsemen are prevented from freely
implementing their informed decisions.
So many high-bracket buyers now depend on those
nickmeisters that an unsatisfactory score can significantly decrease a
horse's value and negate the efforts of horsemen with proven records of
All this has been made possible by the development
of the computer--not so much the powers of the computer as the mystique
of the computer. Consumers of the nicking products believe that if it comes
from a computer, it must be right.
Forgotten is the cautionary phrase of the early
days of computer development: garbage in, garbage out. Forgotten, too,
are some of the most basic rules of traditional statistical research, especially
those that demand adequate sample size and control of variables.
Perhaps most important, basic principles of
genetics have been disregarded.
But the methodology of the nicking gurus is
neither revealed nor examined. Their customers demand answers, not explanations;
"it came out of the computer" is explanation enough.
For reasons to be discussed later, most nicks
involve horses in the third or even fourth generations of the subject horse.
Among the eight sires in the fourth generation of almost every Thoroughbred
are famous horses who can be linked to create that prized nick. Reduce
that to the four sires in the third generation and nicking combinations
are dramatically diminished in number.
But what is the genetic influence of horses
in those generations?
Horses in the fourth generation are, after
all, the great-great-grandparents of the horse that is the subject of the
nicking analysis. We need not challenge the validity of the basic theory
of nicking to inquire as to the chances that that special something will
force its way through the generational barriers to exert its magical influence.
That can be determined simply by applying
the basic law of genetics that says that the maximum chance that a gene
will be passed from one generation to the next is 50 percent. Even the
most dominant of genes make it to the next generation only half the time.
Setting up the most favorable condition possible,
that both sire and dam of the two nicking sires carried that magic gene,
when that law is applied to a four-generation pedigree, the following probabilities
emerge (if the sire and dam of the two nicking sires did not both carry
the gene, the probabilities are one-half those shown below):
Fourth generation to third generation: 100
percent chance that the offspring will receive the gene.
Third generation to second generation: 50
percent of that 100 percent = 50 percent.
Second generation to first generation: 50
percent of 50 percent = 25 percent.
First generation to subject horse: 50 percent
of 25 percent = 12.5 percent.
That 12.5 percent is not the fraction of the
magical gene that the horse will carry. If the gene makes it all the way,
it will be the entire gene. The statistical probability that the horse
will carry that gene at all is 12.5 percent.
That's on one side of the pedigree. To activate
the nick, a gene from the companion sire has to navigate the generational
rapids down the other half of the pedigree, too.
That journey also has a 12.5 percent probability
of success. But the chances that both genes, top and bottom, will survive
the trip are not 12.5 percent. They are 12.5 percent multiplied by 12.5
percent, so the answer is 1.56 percent--one chance in 64.
That is the best possibility, assuming that
the magical element of the nick is carried in just one gene. If it's necessary
for a group of genes to be transmitted, the chances effectively would be
zero. Since a horse carries thousands and thousands of genes, it's hardly
possible that a nick could be made effective by the combining of a single
gene from each of two sires.
Nobody knows how many genes would have to
make that four-generation trip to create a significant combination, but
it's impossible to believe that just one would do the job.
What if just six of those thousands of genes
were required to trigger the nick--three from each of the two nicking sires
four generations back? The chances that just three genes from each horse
would carry across four generations to combine in the subject horse are
0.00038 percent. For comparison, on the same scale, here's the number that
equals one percent: 1.00000.
In other words, the chances effectively are
zero; going four generations back for nicks is a scientific absurdity.
If the nick is restricted to three generations
rather than four, the probabilities are better, but still poor. In the
final calculation 25 percent is multiplied by 25 percent (rather than 12.5
by 12.5) and the answer is 6.25 percent--one chance in 16. Even a two-generation
nick provides no certainty that that gene will be transmitted on both sides
the pedigree. The probability there is just 25 percent (one chance in four).
The highest probability of all, of course, is the
first-generation nick, matching the sire of the prospective foal with the
sire of its dam. That translates to the first generation on one side and
the second generation on the other (again assuming that the sires and dams
of the nicking sires both carried the magic gene). The product of such
a mating has a 50-percent chance of inheriting that prized gene (one chance
The nicking process almost always bogs down
at the first-generation level, though, because few sire/broodmare sire
crosses have produced enough starters to permit the drawing of valid conclusions.
That introduces the major problem of sample
size, which will be discussed in the second part of this series.
One provocative genetic question remains:
What if that magic gene is passed on by a horse or horses other than one
or both of the two nick-linked sires? In the illustrations above,
it is assumed that only those two sires carry the gene that activates the
If the gene is carried by horses other than those
two sires, the presumed nick could be activated even if the gene didn't
descend from either of them. If the recipient of those wild-card genes
raced successfully, the intended nick would get the credit even though
the targeted sires weren't responsible.
The result would be the certification of a
nick as successful when that success was the result of genetic accident
unrelated to the nick.
Part Two: Standards of Research
Let's pretend that you have a friend with
a farm and five stallions. His stallions aren't doing badly, but your friend
wants to book more mares. He thinks that if he cut all the stallions' stud
fees in half, he'd get a lot more mares and make more money. But he wants
to test his plan before he makes the commitment.
So, to test his idea, he places an advertisement
offering just one of the five at a reduced fee. The day after he places
the ad, two runners from that stallion's first crop win stakes races. The
following day, he gets two calls from breeders wanting to book mares to
the horse. Later that same day he decides that his plan has worked and
places another advertisement reducing the stud fees for his other four
Watching all this, you're surely thinking:
a. He doesn't
know whether it was the stud-fee reduction or the two stakes winners that
prompted those people to book their mares. His test was a failure.
He can't interpret the results.
b. Two telephone
calls is not enough on which to base such a decision. There may be no more.
This fictional exercise illustrates two basic
principles of research: stabilizing variables (that's "a" above) and obtaining
large enough samples to support generalizations (that's "b"). It's hard
to believe that anyone would think this make-believe stallion owner acted
sensibly, but customers of the nicking merchants accept--and pay for--the
same kind of bad science.
In the language of science, our friend did
not "stabilize the variables." He couldn't tell whether it was the price
reduction or the stakes winners, or both (or even something else) that
brought in the bookings.
But in the Thoroughbred game, variables abound.
As everyone knows, all mares are not alike. Their female families can be
different, their race records can be different, their produce records can
be different, their conformation can be different, and their temperaments
And that's not all.
Racehorses have different trainers, different
owners, and generally different care and experiences.
As far as nicking analyses are concerned,
the only thing that counts is the mare's sire line. All those other factors
have no bearing on whether the nick is certified as successful or unsuccessful.
All those variables are ignored in nicking analyses.
Even though many different elements contribute to the success or failure
of a racehorse, a nicking analysis is able to consider only the sire lines.
The towering question, then is this: Which
element--or elements--was responsible? The claim that it is always only
the mixture of sire lines is both insupportable and irresponsible.
Flowing alongside that question is another:
How many runners does it take to establish
a nick as either a success or a failure? Putting the question in
another context, how many starters does a new sire have to have before
we are confident that their performance will be typical of the performance
of his future runners?
Would you have to see five runners before
booking your mare to an unproven horse? Ten? Fifteen? Twenty? Maybe
20 aren't enough. Surely five, or even ten, are not.
With that question in mind, consider nicking
information from the Thoroughbred Times Buyer's Guide for the 1994 Keeneland
July yearling sale.
Of the 287 yearlings in the catalog for that
sale, almost half--131--were the result of sire/broodmare sire crosses
without a single runner to the track.
Another 45 had been tried by just one starter,
29 with two, 14 with three, 13 with four, and 11 with five.
That's 243 of the 287 yearlings in the catalog--84.7
percent. Only 15.3 percent were from sire/ brood mare sire nicks that had
been tested by more than five runners.
The Buyer's Guide nicking report then moves
down to the next level--sons of the male-line grandsire crossed with the
sire of the yearling's dam. Even that significant move back into the pedigree
yielded only 168 nicks with more than five starters-- just 58.5 percent
of the yearlings in the catalog.
The third level in the Buyers Guide nicking section,
sons of the male-line grandsire crossed with the grandsire of the yearling's
dam, still found 69 nicks with five or fewer runners and 97 with ten or
Even though they yield larger samples, those
moves backward into the pedigree carry their own hazards, as was detailed
in Part One.
(It is to the credit of the publishers of
the Buyer's Guide that they made no recommendations on the basis of those
statistics. They simply provided the raw information.)
This analysis of those Keeneland yearlings'
pedigrees demonstrates the statistical difficulty of proving the value
of nicks. To summarize, the combination of (a) the impossibility of determining
the factors responsible for a runner's performance and (b) the absence
of racing samples of acceptable size reduces the whole nick-evaluation
process to a kind of interesting parlor game.
It certainly isn't science.
Part Three: Summary and Conclusions
When the current nicking craze was in its early
stages, one of the most acclaimed crosses was the mating of the stallion
Secretariat with mares sired by Crimson Satan. From its first six runners
to the track, that nick produced three stakes winners, two of them multiple
winners of graded races.
Now, several years later, the certification
of the Secretariat/ Crimson Satan cross stands as a textbook example of
the failure of nicking practitioners to observe and follow basic principles
of scientific research.
At the end of 1985, that nick had produced--from
those six starters--the major stakes winners Terlingua ($423,896) and Pancho
Villa ($596,734) as well as Satan's Secretary, winner of unrestricted stakes
at River Downs and Thistledown.
Now, nine years later, that nick has produced 13
starters, and those same three stakes winners. Three stakes winners from
13 starters is an excellent record, but it's not nearly as good as three
stakes winners from six starters.
Putting it in formal terms, a sample of six
runners was too small to permit a valid generalization. In all likelihood,
13 is also too small.
Therein lies one of the serious weaknesses
of the analyses now being generated by nicking merchants all over the country--the
lack of samples of adequate size.
It is instructive to note that, according
to an article in the January issue of The Horseplayer Magazine,
in creating his respected handicapping figures, Andrew Beyer used only
information based on samples of at least 30 horses.
That contrasts sharply with nicking analyses
that routinely accept samples of ten or fewer. The dramatic difference
between Beyer and the nickmeisters is that it was of paramount importance
to Beyer that his numbers be as accurate as possible. He couldn't afford
At the beginning, he was going to the track and
betting his own money to test his system. For that, he required data in
which he could have confidence.
Even more serious than the problem of sample
size is nicking analyses' absolute inability to meet the primary requirement
of scientific method--the stabilizing of all variables except the one being
tested. To qualify as anything more than semi-educated guesses, these analyses
would have to be applied to data in which all elements were identical except
the sire cross being evaluated.
The Secretariat/Crimson Satan nick provides
a dramatic example of the effect of failure to meet that scientific standard.
All Crimson Satan mares are not identical. That fact, as obvious as it
is, is ignored in the certifying of the Secretariat/Crimson Satan nick
as successful and deserving of duplication.
The two best products of that cross, Terlingua and
Pancho Villa, are out of the same mare, Crimson Saint.
Crimson Saint, bred to Secretariat, produced
those two stakes winners plus a Fairplex Park stakes-placed earner of $154,788.
Crimson Saint, bred to Nijinsky 2nd, produced Royal Academy, a Group 1
stakes winner in England and winner of the Grade 1 Breeders' Cup Mile in
the U. S. Crimson Saint, bred to Nijinsky 2nd, produced Encino, third in
the Grade 2 Hollywood Juvenile Championship. Crimson Saint, bred to Alydar,
produced Alydariel, winner of a restricted stakes race at Los Alamitos.
The indisputable fact is that it is impossible
to know whether the Secretariat/Crimson Satan nick was responsible for
Terlingua and Pancho Villa or whether they were simply the result of mating
a high-quality mare to a high-quality stallion. To this day, the Secretariat/Crimson
Satan nick has produced only four stakes horses, and three of them were
out of Crimson Saint.
No reputable scientist could look at this
information and conclude that it proves that Secretariat/Crimson Satan
is a "successful nick." A reputable scientist could conclude only that
nothing has been proved other than that Crimson Saint has been an outstanding
No one needed a computer to figure that out.
But the mare is only one of the elements involved
in the creation and development of a racehorse. The quality of care that
a young horse receives in its early years and the quality of training and
opportunities it receives at the racetrack may be at least as important.
To pretend that all these variables are the
same for all horses is to ignore the basic rules of scientific research
and render worthless any conclusions reached, no matter how attractively
packaged they may be.
Since it is absolutely, permanently impossible
for a nicking analysis to hold constant all the elements that go into making
a racehorse--except those two sire lines--the only conclusion that can
be drawn is that nicking analyses are more likely to mislead than to enlighten.
That is an audacious assertion, considering
that the sale of nicking analyses has become a major industry, with its
practitioners becoming not only wealthy but also respected as the pre-eminent
authorities in the creation of superior racehorses. But this writer is
not the first to note the flaws in the methodology of the nicking merchants.
Dr. Steven A. Roman, creator of the Roman
Dosage Index, made many of the same points in an unpublished paper titled,
"Breeding Theories and Statistics." In the introduction to his lengthy
paper, which he has made available to us, Roman makes this observation
about what he calls "the currently fashionable notion of nicking":
"[It is] an idea that seems attractive on
the surface but upon critical examination
has . . . a questionable scientific basis and may be subject to gross
misuse when accepted at face value."
The problem of sample size noted above is described
by Roman in this way:
The concept of two bloodlines
. . . having a unique compatibility is quite appealing because the decision-
making process is made that much easier. The fundamental problem with individual
nicking patterns . . . is that only rarely are there enough examples to
provide a statistically significant sample size. By statistically significant
we are referring . . . to the established criteria that afford a meaningful
confidence level where the observed pattern has only a small probability
of being a random
event. . . .
Ten or 20 or 30 specific
examples of a cross between [two sires] may still not be sufficient to
establish statistical significance. . . . Bloodline compatibilities are
statistically meaningful only in terms of large populations and to that
extent are useful in assessing trends in bloodlines. Individual nicking
patterns almost never meet the criteria for statistical significance
When samples are too small--as they almost
always are--the nicking analysts either ignore the problem or move back
a generation or more in the pedigree. Regarding this practice, Roman comments:
Using any other representative
of a bloodline as the basis for the compatibility of another member of
that bloodline is a dangerous practice because with each succeeding generation
there is a significant dilution of the genetic relationship between the
two. . . .
Barrera is not Mr. Prospector
even though both are sons of Raise a Native. Wajima is not Secretariat
even though both are sons of Bold Ruler. Anyone who believes a Barrera/Wajima
foal represents the same nicking significance as a Mr. Prospector/ Secretariat
foal had better go back to basics.
Since it is impossible for these nicking analyses
to meet even the most rudimentary criteria for scientific research, why
are they so popular? Roman suggests the answer when he says that the use
of nicking makes the decision-making process easier, but there's more to
it than that.
Most of all, people have embraced nicking
because it offers certainty--or, more correctly, the illusion of
certainty. As presented by its purveyors, nicking offers the certainty
provided by scientifically developed information derived by sophisticated
computerized technology from a vast database of racing and breeding statistics.
Such information is offered not as some individual's
opinion but as the result of impartial, objective scientific research.
Best of all, the answers are simple, clear,
and unambiguous. No one appears to consider the possibility that they also
could be wrong.
As was detailed in the second part of this
series, 131 of the 287 yearlings in the 1994 Keeneland July sale were the
result of sire/broodmare sire crosses untested by even a single runner
to the track. Without a move back at least one generation in the pedigree,
a nicking analysis can produce no honest answer other than "insufficient
Even a move back in the pedigree, to a match
of sons of the yearling's grandsire with the sire of the yearling's dam,
found only 168 nicks tested by more than five starters--only 58.5 percent
of the yearlings in the catalog. Beyond question, any move backward in
the pedigree reduces the validity of the result of the inquiry. The quality
of the conclusion necessarily is degraded.
Yet untold numbers of buyers at the sale paid
significant sums of money for rankings of those yearlings purportedly based
on nicking analyses. In at least one case, those rankings were meticulously
divided into a dozen or more levels, separating the prime prospects from
the not-quite-so prime prospects.
Considering the absence of first-generation
tests of those nicks--and often second-generation as well--the only legitimate
nick ranking for the majority of the yearlings in that sale was "No ranking."
The appropriate response of sellers of nicking information
would have been to refuse to accept payment for ranking those yearlings.
Whether any of them did is not a matter of public record.
If the nicking mania were just an individual
matter affecting only those who believe in its theories, only breeders
and buyers directly involved would be harmed. But the nicking merchants
have gained such power over buyers of racing prospects that sophisticated,
experienced breeders are being forced to modify their mating plans to conform
with the sale recommendations of the nickmeisters.
And though it's never talked about, there
always exists the possibility that the recommendations of a nicking merchant
might be influenced by payments from a stallion owner, sale consignor,
or some other party who might profit from a favorable report.
At present, the industry has no mechanism
of any kind to evaluate the quality of the nickers' products, to verify
their claims, or to expose possible conflicts of interest.
Industry publications meticulously research
claims by stallion owners and sale consignors before accepting their advertisements.
Yet those same publications print advertisements by nicking merchants making
unverified--but always impressive--claims of success.
Whether those who follow the calls of the
nicking sirens deserve our pity or our scorn, the industry as a whole is
the victim of the nicking mania. Despite the claims of its supporters,
the most likely result of their reliance upon nicking as a primary mating
criterion will be the breeding of poorer racehorses than would otherwise
have been the case.
Excluding otherwise desirable stallions from consideration
and including less desirable ones purely on the basis of nicking analyses
inevitably will lead to the producing of less capable runners.
In the long run, the problem will solve itself.
Breeders and buyers who rely on nicking recommendations will be unable
to compete successfully with those who don't. For the present, however,
responsible members of the Thoroughbred community must stand up and defend
legitimate practices rather than tolerate--and sometimes even honor--those
who subvert them.
Editor's note: For
his assistance with the genetic calculations cited above, we wish to thank
Dr. Bo Belhage, associate professor in the Institute of Biochemistry and
Genetics at the University of Copenhagen, Copenhagen, Denmark.
To The Thoroughbred Showcase of the West home page.
Our thanks also go to Dr. Steven A.
Roman of Richmond, Tex., creator of the Roman Dosage Index, for granting
us permission to reprint portions of his unpublished paper analyzing computerized
nicking research practices.