Judgment and inbreeding

I don't think the thoroughbred would exist if there wasn't inbreeding. I think highly inbred horses is just another gamble breeders take.

actually i meant the title to be "Juddmonte" and inbreeding.

Take the point,

What are they inbreeding to these days though? the owner breeder is a dying breed, so is a mare by danehill line, covered by sadlers wells line the ingredient for success?

Juddmonte on the other hand don't care, they want a champion!!

.
https://www.bloodhorse.com/horse-racing/articles/112009/pedigree-analysis-inbreeding-trends


Some food for thought in this article.

Kincsem might have some stats/info/opinions on this topic.

Sorry - didn't see your comments @ 23:07 until after I'd posted.

i_agree_with_nick
Kincsem might have some stats/info/opinions on this topic.
I'm surprised my name was remembered.

That article discusses inbreeding.
What I learned in the spring of 1998 is it is not inbreeding that produces class and speed, it is the correct type of inbreeding(s).
The great majority of inbreeding is two sons (or more) of Northern Dancer, and becoming more common is two sons  (or more) of Danzig (Northern Dancer's son).
I have stats on 159,222 rated horses, and how their inbreeding matches ratings e.g.


In Jan/Feb 2017 I wrote a database program that analyses inbreeding.
Inbreeding of the "right type" improves horse ratings, and more occurrences of the "right type" brings increased improvement.
I compared each of 159,222 ratings with the six generation pedigree of each horse.
The results were informative, and confirmed the theory i.e. good inbreeding produce better horses.
I tested the results statistically.

In simple terms: two sons of a sire (bad, slow plodder); son and daughter of a sire (good); two daughters of a sire (speed no stamina).
But in a pedigree you can have many inbreeding "groups" e.g. Northern Dancer; Hyperion; Blandford; Nearco.
It is almost impossible to plan a pedigree with the good stuff only, some bad inbreeding groups are inevitable.
A pedigree can stand bad inbreeding, if there is also good inbreeding.
My favourite term for two or more sons of a sire in a pedigree - "a recipe for slow".

Two weeks ago I started to re-write the pedigree program.
This time it is eight generation pedigrees (probably no better than six, but "I gots to know").
In December 2017 I bought at auction a broodmare for a planned mating with a stallion picked (and booked) before I bought her.
Three matings this spring with the chosen sire and she is empty.
In the summer there was a bit of vet investigation, laboratory test, and treatment planned before next breeding season.

When I analysed broodmares in a few sales catalogues I came up with the same percentage of mares (+/- 1%) where I could not find ANY good sire match for them ... 87%  ... think about that.
I followed 462 unnamed yearlings sold expensively at a yearling sale a few years back, found their names, racing earnings, win and place records. 
About 15 of 462 had racing earnings that paid for their auction price and a couple of years training.  Perhaps 10 just about broke even or made a small profit.  4 made a profit.

Again this year I plan to compare the Goffs IRE, Tattersalls UK, Arqana FR breeding stock sales Nov/Dec cheap day broodmare to about 600 stallions looking for good matches that might produce a good foal.
Of course if I buy a mare she will be young, inexpensive, by an unfashionable mare, sending her to an inexpensive, unfashionable stallion.
And I want to have the maximum choice of broodmares and stallions to give the maximum number of test-matings e.g. 1,500 inexpensive broodmares with 600 stallions gives 900,000 test-matings.

I want to limit the downside, buying cheaply and using inexpensive sires.
I have no interest in fashionable stallions, black type, sales auction average or median prices, good conformation, good walkers, related to a half-brother who won a race, good pages (catalogue), prices their relatives made.

Frequently I am asked to suggest a stallion for a named mare.
I suggest the best choice imo, but say it might produce nothing more than an average horse.
That advice is ignored and I am then given the name of a stallion and asked would that produce a good horse.
I have no opinion on an individual mares on her own, or stallions on his own.  I am interested in the foal of a stallion and mare..

e-mail me at "my username" at eircom dot net if you want to talk.
Of course my work is on past horses, many running decades ago.

Samples

37,057 Colts - number of inbreeding sire groups
occurrences (i.e. groups); average rating; number of colts
occ    groups    count
0    90.05    876
1    91.44    4974
2    92.42    9889
3    93.74    10803
4    95.24    6826
5    98.07    2662
6    100.15    812
7    104.66    189
8    110.92    24
9    86.00    2


50,335 Geldingss - number of inbreeding sire groups
occurrences (i.e. groups); average rating; number of geldings
occ    groups    count
0    72.89    1052
1    74.35    7372
2    74.88    15017
3    74.59    14778
4    74.26    8439
5    73.65    2857
6    71.91    684
7    70.45    124
8    87.75    8
9    65.00    2
10    76.50    2


Why are the ratings of the colts much better (average 93.79 Vs 74.45)
I looked at this recently and found the colt groups were a little different  - more good inbreeding groups.
The above numbers are further analysed into nine sire and nine dam categories.
In those nine columns I can see which type of inbreeding works and which does not.

And this is the inbreeding type that produces the best results
You can see it is rarer, with 24008 of 37037 colts having 0 (zero) of it.

type    occ    xxxxx    count
colts    0    92.75    24008
colts    1    92.44    4121
colts    2    95.50    4747
colts    3    96.93    2201
colts    4    99.64    1121
colts    5    101.93    532
colts    6    105.67    195
colts    7    106.45    95
colts    8    116.16    19
colts    9    124.55    11
colts    10    112.25    4
colts    11    131.50    2
colts    12    134.00    1


This is the inbreeding type that is common and does not work
type    occ    xxxxx    count
colts    0    93.55    16074
colts    1    93.97    15396
colts    2    93.97    4857
colts    3    93.93    688
colts    4    92.50    42

I do go on a bit.

Good stuff kinscem. For my part, I really like to see good fillies/mares duplicated in a pedigree.

Very ionteresting, kincsem. Keep it coming.

Kincsem, I don't understand the two final tables.

Presumably xxxxx refers to the inbreeding type? But it is the same in both tables.

What is xxxxx? Does that mean inbred within a particular generation or a specific position within the pedigree?

i_agree_with_nick
Kincsem, I don't understand the two final tables.
Presumably xxxxx refers to the inbreeding type?

I sent you a message with an explanation.

There are 18 inbreeding group types in the pedigree analysis program I wrote in Jan/Feb 2017.*
*the duplication must be on both sides of the pedigree, part in the sire side of the foal, part in the dam side

Back in 2000 I wrote an earlier version that took 18 months .... poor programming skills, employed in a very busy job, getting up at 6am, playing poker at night.
The earlier version was not much use, but some use.
In 2006 I was lucky to get early retirement at age 56.  The company closed 18 months later.

It was obvious one reason the earlier program was a poor predictor.
My pedigree database was group horses and their ancestors only (or the races that preceded the Group/pattern/black type).
I did not have any, or many, bad horses in the data to provide a contrast.

I have an intense dislike for black type, imo almost useless.
A test showed only 11% of group race winners (IRE, GB, FR, GER, ITY, USA (Grade 1 only)) had dams who won Group races. 
That should tell people who go by sales catalogue black type that they are missing 89% of the Group winner producing dams.

What would be useful when testing a pedigree program is thousands of poor horses, and their pedigrees.
They do not have black type, but they do have ratings.
You need the full spectrum: champions, the group winners, listed, handicappers, non-winners, non-placed.
The program, when written, would be tested against ratings.


PROGRAM RESULTS
The program generates a six generation pedigree of a horse (2+4+8+16+32+64=126 ancestors), counts the inbreeding groups by type, and outputs these to a results file, then moves to the next horse in the load file and repeats the analysis.

groups .... a group is in inbreeding duplication group e.g. two sons and a daughter of Northern Dancer
off_m .... the number of male offspring (sons) of duplication groups
off_f .... as above for females (daughters)
s_mf .... a duplicated sire (s) producing a son (m) and daughter (f)
s_mm .... a duplicated sire (s) producing two sons (mm)
s_ff.... a duplicated sire (s) producing two daughters (ff)
s_mmff.... a duplicated sire (s) producing sons (m) and daughters (f), three or more e.g. mmf, mff, mfmf, fffm, mmfm, mfmfmf, mmmmmf
s_mmmm... a duplicated sire (s) producing three or more sons (m) e.g. mmm, mmmm, mmmmm
s_ffff .... a duplicated sire (s) producing three or more daughters (f) e.g. fff, ffff, ffffff
s_one ... one sided duplication groups i.e. in the sire side only, or the dam side only
s_supp .... support horses i.e. the sire or dam of the duplication group sire/dam i.e. if Sadler's Wells is a duplication group, then Northern Dancer, his sire, elsewhere in the pedigree but not in a duplication group is "support".
s_sibl ... full-siblings - two full brothers, two full sisters, a full sister and brother e.g. Sadler's Wells and Fairy King are full-sibling brothers (both by Northern Dancer out of Fairy Bridge).

Then do the above again for duplication groups of dams/mares
e.g. d_mf; d_mm; d_ff; d_mmff; d_mmmm; d_ffff; d_one; d_supp; d_sibl

occ in the above charts is "occurrences" - the number of groups.
The first xxxxx is the s_sibl information for 37057 colts only in my test data. (not the 50335 geldings or 71830 fillies, their numbers are flatline, not as impressive)
It indicates that more full-sibling groups give increases in ratings (for the colts).
The average rating for the colts was 93.79, the geldings 74.45, the fillies 74.43.

The second xxxxx above is s_mm ... sire duplication groups producing two male offspring (sons).


You must remember that the average for every colt inbreeding type (s_mf, s_mm, s_sibl, s_supp) is always 93.79.
A difficulty is separating the effect of the different inbreedings.
A horse could have 1 x s_mf; 3 x s_mm; 1 x d_ff; 2 x s_supp.
How much of his rating is helped by one type of inbreeding, and hindered by the other?

The results of the Jan/Feb 2017 program were tested statistically and gave positive connections between some of the inbreeding types and ratings.
From memory the program processed* 40 pedigrees a second on my fast PC (I program on the slow PC).
*load a horse, generated a six generation pedigree, extract the inbreeding duplications, the sex of the offspring, output to a results file .... then load the next horse)

The revised version I am trying to write now will have other (extra) analysis.
I will use an 8 generation pedigree for each horse (510 ancestors).
I a sales catalogue you get 3 generations (14 ancestors).

Thank you.

I have just sent you a PM.

kincsem
21 Oct 18 11:31

I have an intense dislike for black type, imo almost useless.
A test showed only 11% of group race winners (IRE, GB, FR, GER, ITY, USA (Grade 1 only)) had dams who won Group races.
That should tell people who go by sales catalogue black type that they are missing 89% of the Group winner producing dams.


Yes but I guess this stat would be more meaningful if we knew what percentage of all runners in Group races had dams who had won Group races.