Northern Dancer–Regression to the Mean

NOTE: I have been castigated for using the word “heritability,” as you may have noticed if you look at the comments on this blog. I have agreed not to use the word again, not because I am convinced I did so incorrectly, but because the word itself is unimportant. I will use the word “regression” in its place, as in regression to the mean, the norm, the average, take your pick of a word for “middle.”

——————————————————————————————-

Northern Dancer was one of the greatest sires of the 20th century. As such he was bred to the best available mares of his time (particularly after his first few crops had raced).

Did ND sire 100% stakes winners from foals???? Of course not. Did ND sire 50% stakes winners from foals???? Of course not.

ND did sire 23% stakes winners (147) from foals (646). That is not a record, but it is very close. And it was not only 23% stakes winners from foals that made ND such a great sire, but also the fact that so many of those stakes winners were champions and G1 winners.

So the greatest sires bred to the best mares available to them do not sire 100% stakes winners. Nor even 50% stakes winners. The actual figure of 23% is considered remarkably good. It is about 667% better than the overall breed norm of 3% stakes winners from foals. The expectation for a dual classic winner such as ND might have been about 10% stakes winners from foals at that time (now it is much lower). So if you look at it that way, you could say that ND outperformed expectations by at least 130%. I say “at least” because so many of his stakes winners were champions and G1 winners.

What is interesting to me is the difference between ND’s numbers as a sire in the first generation and his numbers as a sire in the second generation. Some actual numbers are available on ND and other sires as broodmare sires (second generation). I will quote two sets of them.

At the end of 1984 ND was represented by 54 stakes winners from 504 foals as a broodmare sire (10.71%). If “influence” declined by ONLY 50% per generation, that number would be expected to be 13%. As a sire he had 23% stakes winners from foals. That is an improvement of 20% in absolute terms over the breed norm. Because ND is now one generation farther back (in the second as opposed to the first generation), we should expect to see an improvement of 10% in the second generation; 3% (the norm) plus 10% (the expected improvement) equals 13%.

ND actually came close to that 13% expectation. He got 77% of the way there (and I will term this “regression”). I arrive at that figure of 77% by dividing 7.71% (10.71% minus 3%, or the amount of actual improvement) by 10% (the expected amount of improvement if “influence” declined by ONLY 50% per generation).

Now let us look at ND’s current figures as a broodmare sire. As of today, according to TJC numbers available on Equineline, ND has been represented by 227 stakes winners from 2,728 foals as a broodmare sire (8.32%).

So his performance as a broodmare sire has slipped from 10.71% stakes winners from foals at the end of 1984 to 8.32% stakes winners from foals today. What are we to make of this????

I am not surprised in the least. It has been my observation that most statistics on sires and broodmare sires tend to deteriorate over time. And 8.32% is still not bad at all. It represents a regression of about 53% (compared to 77% at the end of 1984, same calculations as above).

What about ND as a sire of sires (second generation)???? For every foal that ND had as a broodmare sire, he probably had at least ten foals as a sire of sires. So if he had 2,728 foals as a broodmare sire, he probably had at least 30,000 foals as a sire of sires. Maybe even 50,000. Maybe even 100,000 worldwide. Who knows???? And if you want to assess ND’s overall contribution in the second generation, his numbers as a sires of sires dwarf his numbers as a broodmare sire.

The problem is that nobody knows what the actual numbers are because no one has actually compiled such numbers on sires of sires.

Except for yours truly of course. I ran these numbers on ND as a sire of sires among all North American-bred foals of 1983 back in the early 1990s. I came up with approximately 6.5% stakes winners from foals for ND as a sire of sires. I do not have the exact numbers, for which I apologize. I don’t think we even published them at the time (no one was particularly interested). But 6.5% is definitely the number that sticks in my mind.

Humor me and accept 6.5% as a ballpark figure. If ND had 6.5% stakes winners from foals as a sire of sires, then his regression as a sire of sires was about 35%. If “influence” declines by ONLY 50% per generation, he should have had about 13% stakes winners from foals as a sire of sires (see same section above on broodmare sires). He actually had about 6.5%.

I arrive at that figure of 35% by dividing 3.5% (6.5% minus 3%; the amount of observed improvement over the norm of 3%) by 10% (13% minus 3%, the amount of expected improvement). Add in ND’s numbers as a broodmare sire, and the figure drifts up to 37% or so, still well below the 100% expected if “influence” declines by ONLY 50% per generation.

I have thought of another source of statistics on this matter. See Chapter 9, page 176, Racehorse Breeding Theories. Among all North American-bred foals of 1995-1997, Nearctic had 237 stakes winners from 4,103 foals (5.78%) at P1 in the third generation.

Northern Dancer at P1 in the second generation (sire of sires) is AT LEAST 90% of Nearctic at P1 in the third generation (sire of sires of sires). SO ND accounts for at least 90% of this figure. Which means that ND’s results at P1 in the second generation are not going to be very far off from Nearctic’s results at P1 in the third generation.

That 5.78% stakes winners from foals was counting graded stakes winners twice. So the actual figure is well below 5%, which in turn is well below the 6.5% I remember for all North American-bred foals of 1983.

At any rate, the important point is that both 6.5% and 5% are WELL BELOW the expected figure of 13% if “influence” declines by ONLY 50% per generation.

In my next post I will detail some other numbers on the regression of some other sires and broodmare sires.

Advertisements
This entry was posted in Uncategorized and tagged . Bookmark the permalink.

One Response to Northern Dancer–Regression to the Mean

  1. Byron Rogers says:

    Boojum,

    Interesting post. It speaks to much of Bertrand Langlois’s work where he showed that heritability of racing speed at ~30% so as each generation goes by the ancestor named is less influential, i.e If the heritability of the trait is ~30% then the sire and dam is responsible for approx ~15% each, the grandparents ~7.5% each and the great grandparents ~3.25%. Once you get past there you pretty much are just talking names on the page as far as their influence on the racing ability of a horse.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s