Revista Científica UDO Agrícola Volumen 8.
Número 1. Año 2008. Páginas: 127-131
Quantitative study for
race times in thoroughbreds
on dirt and turf tracks in Brazil
Estudio cuantitativo de los tiempos en las carreras de
purasangre en pistas de tierra y césped en el Brasil
Marcilio Dias Silveira Da
MOTA
1 and Daniel Madureira
GOUVEIA FERREIRA2
1Department of Animal
Breeding and Nutrition, School of Veterinary Medicine and Animal Science, São
Paulo State University C.P.560, CEP
18618-000, Botucatu/SP, Brazil and 2University
of Trás-os-Montes e Alto D’ouro,
Vila Real, Portugal.
Email: mdsmota@fca.unesp.br Corresponding author
|
Received: 02/25/2008 |
First reviewing ending:
04/30/2008 |
|
First review received: 06/16/2008 |
Accepted: 06/22/2008 |
ABSTRACT
This study was conducted
to estimate the genetic parameters
for race times on turf and dirt tracks of Thoroughbred race horses. Data used were recorded
by the Turftotal
Ltd. for 343,419 racing
performance of 26,713 animals, from
January 1992 to January 2003. The model used in analysis
included random animal and permanent environmental effects, and age, post position
at start, sex and race as fixed effects. The variance and covariance components and the breeding value
were estimated using the MTGSAM software. Heritability estimates were 0.29 for time on dirt track
and 0.25 for time on turf track, indicating a moderate association between the animals’
breeding values and their phenotypic values. Although genetic and environmental variances were smaller in turf tracks, their repeatability was equal to
that of dirt (0.56), in terms of the highest
estimated phenotypic variance for the
latter type of track. The genetic
correlation between times on different tracks
was high (0.70). Considering the mean breeding value of the progeny of 465 stallions with 10 or more offspring, Spearman’s correlation was 0.80, indicating that most Thoroughbred
stallions produce offspring
suited to both dirt and turf racing tracks.
Key words: Genetic
parameters, horserace, race time, turf and dirt tracks
RESUMEN
Este estudio fue conducido para estimar los parámetros
genéticos de los tiempos de carrera en pistas de tierra y césped de caballos
purasangres. Los datos utilizados fueron registrados por Turftotal
Ltd. para 343.419 desempeños en carreras de 26.713 animales, entre enero de
1992 y enero de 2003. El modelo utilizado en el análisis incluyó animales
aleatorios y efectos ambientales permanentes, y la edad, puesto en la posición
de partida, sexo y raza como efectos fijos. Los componentes de varianza y covarianza
y el valor genético se estimaron usando el programa MTGSAM. Los estimados de heredabilidad fueron 0,29 para el tiempo en pista de tierra
y 0,25 para el tiempo en pista de césped, indicando una asociación moderada
entre los valores genéticos de los animales y sus valores fenotípicos. A pesar
de que las varianzas genética y ambiental fueron menores en las pistas de
césped, su repetibilidad fue igual a aquella de las
pistas de tierra (0,56), en términos de la más alta varianza fenotípica
estimada para el último tipo de pista. La correlación genética entre los
tiempos en las diferentes pistas fue alta (0,70). Considerando el valor
genético promedio de la progenie de 465 sementales con 10 o más hijos, la
correlación de Spearman fue 0,80, indicando que la
mayoría de los sementales purasangres producen descendientes adecuados para
ambas pistas de carrera (tierra y césped).
Palabras
clave: parámetros genéticos, carreras
de caballo, tiempo de carrera, pistas de tierra y césped.
INTRODUCTION
Thoroughbred
racing in Brazil occurs on both dirt (approximately 62%) and turf tracks (Turftotal, 2005). Although racing on dirt tracks are more
numerous, those of greatest standing and economically most important occur on
turf. Thus, unlike in some European countries, where the animals normally race
on turf, and in the United States, where there is a preference for dirt tracks,
in Brazil, breeders usually seek horses with good performances on both types of
surface.
Genetic
parameter estimates for race times on these surface types have been shown to be
variable in the literature. While Moritsu et al. (1998) reported higher
heritability estimates for race times on turf, Oki et al. (1995) reported the opposite. Genetic correlations between
race times at different distances on dirt and turf track have been moderate and
favorable (Oki et al. 1995). However,
in Brazil, the genetic variability for race times on both these track types is
unknown, as is the existing relationship between them, and these are fundamental
steps for the elaboration of consistent programs for genetic improvement.
In
this context, the study was conducted to estimate the genetic parameters for
race times in Thoroughbreds on dirt and turf tracks, with the aim of assisting
selection programs in animals of this breed.
MATERIAL AND
METHODS
The
data used in this study were provided by the company Turftotal
Ltd. and comprised of 343,419 race time
observations for 26,713 animals (14,073 male and 12,640 female) of the
Thoroughbred, over 12 years (from January 1992 to January 2003), on the
principal Brazilian race tracks (Cidade Jardim - São Paulo State, Gávea -
Rio de Janeiro, Tarumã - Paraná and Cristal - Rio
Grande do Sul). The total number of animals in the
relationship matrix was 37,444. Table 1 illustrates the descriptive analysis of this information.
|
Table 1. Number of observations, number of races,
number of animals, mean number of animals per race and mean, minimum and
maximum values for distances, according to track type. |
|||||||
|
Track |
Nº
of observations |
Nº
of races |
Nº
of animals |
Mean
animals/race |
Distance
(m) |
||
|
Mean |
Min. |
Max. |
|||||
|
Dirt |
226.917 |
28.990 |
24.969 |
8.7 |
1319 |
600 |
3200 |
|
Turf |
116.502 |
13.045 |
20.960 |
9.8 |
1379 |
1000 |
3500 |
Each
race included a minimum of four animals and the mean number of start per horse
was 10.7 (range 1 to 121). With regard to the 26,713 animals evaluated, 5,753
only ran on dirt tracks, 1,744 only ran on turf and the rest (19,216) ran on
both types of surface. The first race was on dirt surface for 65.3% of the
animals.
The
SAS statistical software (1999) was used to set up the files, to determine
consistency and to perform descriptive analysis of the traits. The model used
to estimate the (co)variance components necessary to obtain the genetic
parameters (heritability, repeatability and correlations) of the traits was the
same used by Mota et
al. (2005), but in a two-trait animal model, and included the random animal
and permanent effects, and the fixed effects of post position at start (1 to ≥
11), age (3 years or younger, 4, 5 and
older than 5 years), sex (male or female) and race (1 to 42.035).
In matrix terms:
where:
y = vector of time records on turf and dirt tracks;
β = vector of fixed effects of
race, sex, age and post position at start, as associated with records in y by
X;
α = vector of the additive genetic
effects as associated with the records in
y by Z;
p = vector of the
permanent environmental effects as associated with the records in y by Z;
e = vector of
residual effects;
X and Z = incidence matrices relating a
particular record to a particular animal.
The program used to obtain (co)variance components and breeding values of the animals was
the MTGSAM (Multiple - Trait Gibbs Sampler
for Animal Models), described by Van Tassel and Van Vleck (1995). Inferences regarding parameter dispersion were realized using
the distributions obtained a posteriori via the Gibbs sampler.
For the additive
genetic, permanent environmental variances and
residual variances non-informative
(Flat) a priori distributions were used. The
Fortran Gibanal program, version 2.4 (VanKaam, 1998), was used to
analyze the Gibbs series, with the aim of defining
the burn-in period, series parameter spacing and total number of samples to be used.
The Gibbs sampling scheme considered the series size of 505,000, with a burn-in of 5,000 and spacing to 500, resulting in 1000 available samples for a posteriori distribution evaluation.
The heritabilities were
estimated from the ratio between the additive genetic
variance (σ2A) and the phenotypic variance (σ2P), and repeatability was calculated by dividing
the sum of the additive genetic variance and the environmental variance (σ2EP)
by the phenotypic
variance.
To
calculate the efficiency of “n” records in relation to a single record, the
expression described by Cardellino and Rovira (1989) was used:
En=
E1
where:
En =
efficiency when “n” records are realized;
E1 =
efficiency when only 1 record is realized;
n = number of records
realized;
t = repeatability
of the trait analyzed.
An
evaluation of rank correlation was made between the stallion’s breeding values
based on the progeny’s racing time (s) on two kinds of racing tracks.
RESULTS AND DISCUSSION
Table 2
shows the components of variance, heritability and repeatability of the studied
traits.
|
Table 2. Mean and respective standard errors (SD),
mode, highest a posteriori density intervals with 90% probability
(HPD90%), minimum (Min.) and maximum (Max.) values of the genetic additive (s2A), permanent
environmental (s2EP), residual (s2R) and phenotypic
(s2P) variances,
heritability (h2) and repeatability (t) for race times on dirt and
turf tracks.* |
||||||
|
Track
type / Item |
Mean |
SD |
Mode |
HPD. 90% |
Min. |
Max. |
|
Dirt
Track |
|
|
|
|
|
|
|
s2A |
0.93 |
0.03 |
0.93 |
0.87 to 0.99 |
0.82 |
1.09 |
|
s2EP |
0.86 |
0.025 |
0.86 |
0.81 to 0.91 |
0.77 |
0.94 |
|
s2R |
1.43 |
0.004 |
1.43 |
1.43 to 1.44 |
1.42 |
1.45 |
|
s2P |
3.23 |
0.03 |
3.20 |
3.19 to 3.28 |
3.16 |
3.32 |
|
h2 |
0.29 |
0.01 |
0.29 |
0.27 to 0.31 |
0.26 |
0.33 |
|
t |
0.56 |
0.004 |
0.56 |
0.55 to 0.56 |
0.54 |
0.57 |
|
Turf Track |
|
|
|
|
|
|
|
s2A |
0.61 |
0.03 |
0.61 |
0.56 to 0.66 |
0.52 |
0.71 |
|
s2EP |
0.72 |
0.02 |
0.72 |
0.68 to 0.76 |
0.66 |
0.79 |
|
s2R |
1.07 |
0.004 |
1.07 |
1.06 to 1.08 |
1.05 |
1.08 |
|
s2P |
2.40 |
0.02 |
2.41 |
2.37 to 2.44 |
2.34 |
2.46 |
|
h2 |
0.25 |
0.05 |
0.25 |
0.23 to 0.27 |
0.22 |
0.28 |
|
t |
0.56 |
0.005 |
0.56 |
0.55 to 0.56 |
0.54 |
0.57 |
|
*
the convergence criteria used was 10-9 |
||||||
All
estimated variances were higher for dirt tracks, in relation to turf, where the
greatest difference was verified between the phenotypic variances. The estimates
for heritability were of medium magnitude both for dirt (0.29) and for turf
(0.25), indicating a moderate association between the animals’ breeding values
and their phenotypic values (performances), and the possibility of a slightly
less effective response to selection if applied to the latter trait (Figure 1).
Oki et al. (1995), evaluating race time at five
distances on dirt and six on turf, also reported a higher heritability for the
first type of track surface (mean equal to 0.16) in relation to the second
(mean equal to 0.13). However, all estimated variances (additive genetic,
permanent environmental, residual and phenotypic) by these authors were
inferior to those found in this study. In contrast, Moritsu
et al. (1998), working with rating
scores for Thoroughbreds in Japan, estimated greater heritability for turf tracks
(0.29) in relation to those with dirt (0.18). In Arabian horses, Ekiz et al.
(2005), studying race times at different distances on dirt and turf tracks,
estimated heritability varying from 0.17 to 0.30, amplitude which include the
values observed in the present study.
Although
the genetic additive and permanent environmental variances were lower on turf,
the repeatability was equal to that for dirt (0.56) in terms of the highest estimated
phenotypic variance for the latter type of track. This means that the relation
intensity between time measurements taken over an individual animal’s lifetime
is similar for both types of track surface, indicating that, independent of the
track type, a single performance of an animal is a moderate indicator of its
production capacity. The repeatability estimates agree with that reported by Tolley et al. (1983), Saastamoinen and Ojala (1991) in
Trotters, Oki et al. (1995) in
Thoroughbred and Villela et al. (2002) and Corrêa and Mota (2007) in Quarter Horse 
a mean value of 0.56. In contrast, Ekiz et al.
(2005) and Mota et
al. (2005) observed inferior repeatabilities in
Arabian (0.29 to 0.46) and Thoroughbreds (mean value of 0.32) horses.
Considering
that the mean number of starts per animal was 10.7, a 29% higher efficiency (in
relation to a single start) would be achieved if breeders considered this mean
starting number before culling the animals.
The genetic correlation
between race times in different track types was high
(0.70), showing that animals genetically superior for turf racing tend to have
higher genetic values for dirt
as well, although correlations less than 0.80, according to Robertson (1959), indicate the presence of genotype x environmental interaction, that is, the animals’
genotypes express themselves in a distinct way depending on
the track surface. Considering that the genetic
and permanent environmental
correlations between these two characteristics
were high (0.70 and 0.67, respectively) and that the residual correlation was slightly negative
(-0.03), resulting in a moderate
phenotypic correlation
(0.38), performances in both types
of track surface fundamentally depend on the action
of the same group of genes and the effects of the permanent environment (breeder, trainer, injuries,
etc.). With regard to the mean breeding
value of the progeny of 465 stallions with ten or more offspring, for dirt and turf tracks, Spearman’s correlation was 0.80, indicating that most Thoroughbred
stallions in Brazil produce
offspring suited to both dirt
and turf racing tracks
(Figure 2). This result is in agreement with the report
by Moritsu et al. (1998) in Thoroughbreds
in Japan, although the authors found
that Spearman’s correlation was lower (0.50), evaluating progeny of 116 stallions.
The
correlation between predicted mean breeding values for race times on turf and on
dirt was high, although there was considerable variation in animal
classification depending on the criterion used. Figure 3 presents the mean
breeding values for race times on turf tracks when the selection was performed
using the predicted mean breeding value for time on turf surfaces as criterion,
and when using predicted breeding value on dirt tracks. There was a reduction
in the response to selection in time for turf tracks if the selection was based
on times for dirt tracks. This difference, depending on the fraction selected,
reached more than 1 tenth of a second in terms of the mean breeding value.
CONCLUSIONS
The results
found in this study indicate that there is sufficient additive genetic
variability for race time selection in Thoroughbred horses in Brazil, on both
dirt and turf tracks, with a slight superiority for the former.
Additionally,
although Thoroughbred horse breeders desire animals with superior performance
on both track types (dirt and turf) and that this, in general, occurs, it is
important that judicious appreciation be made during the selection of
genetically superior animals, for both track types, principally with those that
form the top ten, with the aim of maximizing genetic gains.
LITERATURE CITED
Cardellino,
R. and J. Rovira. 1989. Mejoramiento genético animal, Ed. Agropecuária,
Hemisfério Sur, Montevideo, Uruguai,
253 p.
Corrêa,
M. J. M. and M. D. S. Mota. 2007. Genetic evaluation of performance traits in Brazilian Quarter Horse. Journal of Applied
Genetics 48:145-151.
Ekiz, B.; O. Koçak
and H. Demir. 2005. Estimates of genetic parameters
for racing performances of Arabian horses. Turkey Journal of
Veterinary and Animal Science 29: 543-549.
Moritsu, Y.; A. Terai and T. Tashiro. 1998. Relationship between sire Breeding Values for
the Rating Score on Turf and Dirt Racing Tracks in Thoroughbred Racehorses.
Journal of Equine Science 9: 82-92.
Mota, M. D. S.; A. R. Abrahão and H. N. Oliveira. 2005. Genetic and environmental
parameters for racing time at different distances in Brazilian Thoroughbreds.
Journal of Animal Breeding and Genetics, 122: 393-399.
Oki, H.; Y. Sasaki and R. L. Willham. 1995. Genetic parameter estimates for racing
time by restricted maximum likelihood in the Thoroughbred horse of Japan.
Journal of Animal Breeding and Genetics, 112: 146-150.
Robertson, A. 1959. The
sampling variance of the genetic correlation coefficient. Biometrics,
15: 469-472.
Saastamoinen, M. T. and
M. J. Ojala. 1991. Influence of birth-month on age at first start and
racing performance in young trotters. Acta Agriculture Scandinavia, 41: 437-445.
SAS - Statistical Analysis System 1999: SAS User’s guide: Statistical,
version. 8. Cary: SAS Institute Inc, 956 p.
Tolley,
E.A.; D. R. Notter and T. J. Marlowe. 1983. Heritability
and repeatability of speed for 2 and 3 year old Standardbred
racehorses. Journal of Animal Science, 56: 12941305.
Turftotal Ltd., 2005. Thoroughbred data
control company. Retrieved January 15 2005, from www.turftotal.com.br
(in Portuguese).
VanKaam, J. B. C. H. M. 1998. Gibanal program.
Retrieved November 28 2005, from www.student.Wau.nl/~janthijs/ breedingsite/edagibanal.html.
Van Tassel, C. P. and L. D. Van Vleck. 1995. A manual for use of MTGSAM. A set of FORTRAM
programs to apply Gibbs sampling to animal models for variance component
estimation [DRAFT] U.S. Department of agriculture, Agriculture. Agricultural Research Service.
Villela, L. C. V.; M.
D. S. Mota and H. N. Oliveira. 2002. Genetic parameters of racing performance traits
of Quarter horses in Brazil. Journal of Animal Breeding and Genetics, 119:
229-234.
Página diseñada por Prof. Jesús Rafael Méndez Natera
TABLA DE CONTENIDO DE LA REVISTA CIENTÍFICA UDO
AGRÍCOLA
