griff wrote:Suggest everyone go back to page 1 and read posy #2,
griff
And I
STRONGLY suggest they don't bother. Jack Werk may well be a decent fellow and good pedigree analyst (I don't really know enough about him to know either way), but nonetheless Werk Nicks are too flawed to be of any value to horse breeders to the best of my knowledge (and I’m almost certain I've spent more time thinking about this issue than the vast majority of people who have voiced opinions on this subject).
Besides Michael's point about lack of discrimination between Graded and Non-Graded Stakes; there is, in fact, the additional and utterly fatal flaw of the system not taking into account the Geographic Concentration of the two Male Lines you imput to produce the score.
Anyways, Byron Rogers (the Stallion Manager of a little outfit called Taylor Made Farm) put it better than I:
brogers
Posted: Tue Jun 07, 2005 1:04 pm Post subject: Werk nick ratings aren't calculated properly
--------------------------------------------------------------------------------
What you should all understand is that Werk Nick Ratings make no allowance for opportunity.
Werk Nick Ratings are achieved by using their own database which is restricted to approximately 25,000 stakes winners worldwide. The program also does not include restricted stakes winners, even though they are often superior to minor unrestricted stakes winners. For example the likes of Archers Bay, a Grade One winner in Canada is not in the population because the grade one race he won was restricted to Canadian breds even though he was a much better horse than many of the listed stakes winners in their database. A formula below (which was published in the Australian Bloodhorse Review so has entered the public domain) is then utilized to calculate the nick.
No. of stakes winners for the nick
Divided by
No. of stakes winners for the sire line
Over
No. of stakes winners for the broodmare sire line
Divided by
No of stakes winners in the database
Anyone with even a basic understanding of mathematics will appreciate that they are comparing apples to oranges. There is no factor considering the number of times that the cross has actually been tried.
The key problem here is that the correction for opportunity is only hypothetical. It assumes that any mare in the population could be bred to any sire in the population, whereas geographic and other factors ensure that this is not so.
An excellent example to use here is that of the cross of leading South American sire, Southern Halo with mares by Logical. The Werk Nick rating has the Southern Halo/Logical cross performing 352 times better than opportunity resulting A++ rating.
It is true that of the 121 stakes winners sired by Southern Halo, at least 26 (21.4%) are out of mares by Logical. What is not evident from this statistic, however, is that Southern Halo and Logical stood at the same farm. As a result, Southern Halo covered a far higher number of Logical mares than theoretical opportunity would dictate. In fact, of 796 known foals by Southern Halo, 76 (9.9%) are out of mares by Logical. Applying a straight opportunity based formula we find that the Southern Halo/Logical cross is, in reality, outperforming opportunity by a factor of 2.61, rather than the 352 suggest by Werk.
Despite Joe Estes, nicks are valid guidelines to use when assessing a potential mating but the way Werk does it is just mathematically flawed and only comes about because they don't have a complete database to factor opportunity.
(Please See:
http://www.pedigreequery.com/forum/view ... werk#35614 .)
It may be suspected that I'm just criticizing Werk Nicks to try and get people to use the Venture GeoScore (which, while certainly not perfect as yet, at least lacks Werk's glaring flaws of not taking into account Stakes Quality and the Geographical Opportunity for a cross). But in contradiction of that insinuation I'd like to note that if Horse Breeders simply stopped using Werk Nicks without even switching to the VGS; I'd be much happier than I am now, in a World of breeding decisions driven by the statistical techniques so fundamentally flawed as to be worthless.
-llbean
