GENETICS OF RACING PERFORMANCE IN THE AMERICAN QUARTER HORSE: I. DESCRIPTION OF THE DATA 1'2

Similar documents
GENETICS OF RACING PERFORMANCE IN THE AMERICAN QUARTER HORSE: II. ADJUSTMENT FACTORS AND CONTEMPORARY GROUPS 1'2

Genetics of racing performance in the American Quarter Horse

Genetic correlations between racing performance at different racing distances in Thoroughbreds and Arab horses

The Irish Sport Horse Studbook Genetic Evaluation Report 2015

Salers 2015 Sire Summary Definitions

Explore the basis for including competition traits in the genetic evaluation of the Icelandic horse

Genetic Improvement for. Auxiliary Traits in Canada


JUVENILE GENETIC EVALUATION AND SELECTION FOR DAIRY GOATS

Access to the published version may require journal subscription. Published with permission from: Cambridge University Press

Contemporary Grouping for Beef Cattle Genetic Evaluation

Estimating breeding value

A N E X P L O R AT I O N W I T H N E W Y O R K C I T Y TA X I D ATA S E T

Performance Data Reporting and Contemporary Grouping For Beef Cattle Genetic Evaluation

EXPECTED PROGENY DIFFERENCES

A Tobit-Model for breeding value evaluation in German trotters

Implications of across-studbook genetic correlations between linear traits for sport horse breeding

September Population analysis of the Finnish Spitz breed

Test Results Summary. Owner: Darby Dan Farm 8 th Nov Distance Plus (USA) v1.0. Distance Plus (SAF) v1.0. Distance Plus (ANZ) v1.

-- Results and Discussion

A Hedonic Price Analysis of Risk Preferences in Yearling Thoroughbred Buyers. Xiurui Iris Cui

Analysis of Estimated Breeding Values for Marble Score, Carcase Weight and the Terminal Carcase Index

September Population analysis of the Portuguese Water Dog breed

Bull Buyer s Guide. $3000 Purchase Price of New Bull Salvage Value of Old Bull (1900 lbs. X 1.10/lb.) $ 910 Net Cost of New Bull

The Aging Curve(s) Jerry Meyer, Central Maryland YMCA Masters (CMYM)

September Population analysis of the Dogue De Bordeaux breed

WHAT CAN WE LEARN FROM COMPETITION ANALYSIS AT THE 1999 PAN PACIFIC SWIMMING CHAMPIONSHIPS?

Genetic Analysis of Eventing Data in the Swedish Warmblood Population

EFFECT OF THE LINEAL AND FAMILY BELONGING OF BROODMARES FROM THE SHAGYA BREED WITH RESPECT TO THE HEIGHT AT WITHERS AND CANNON GIRTH

CALF PERFORMANCE AND COW WEIGHT AND CONDITION FOR COWS SIRED BY HIGH AND LOW MILK EPD ANGUS AND POLLED HEREFORD BULLS

Variation of Nordic Classic Ski Characteristics from Norwegian national team athletes

Value of Black Hereford Registration

APPROACH RUN VELOCITIES OF FEMALE POLE VAULTERS

Running head: DATA ANALYSIS AND INTERPRETATION 1

Competitive Performance of Elite Olympic-Distance Triathletes: Reliability and Smallest Worthwhile Enhancement

MICROPHONE WIND SPEED LIMITS DURING WIND FARM NOISE MEASUREMENTS

Equine Injury Database Update and Call for More Data

2015 IOWA FACT BOOK. A statistical guide to the Thoroughbred industry in Iowa. Prepared by The Jockey Club

Value of Black Hereford Registration

Maximizing genetic progress in the new age of genomics

Evaluating the Influence of R3 Treatments on Fishing License Sales in Pennsylvania

Hair Shedding Scores Relating to Maternal Traits and Productivity in Beef Cattle. An Undergraduate Honors Thesis in the. Animal Science Department

Compression Study: City, State. City Convention & Visitors Bureau. Prepared for

A SURVEY OF 1997 COLORADO ANGLERS AND THEIR WILLINGNESS TO PAY INCREASED LICENSE FEES

Regression to the Mean at The Masters Golf Tournament A comparative analysis of regression to the mean on the PGA tour and at the Masters Tournament

Bicycle and Pedestrian Level of Service Evaluation

COMMERCIAL BEEF SIRE SELECTION

Maturity Pattern in Simmentals

ABSTRACT AUTHOR. Kinematic Analysis of the Women's 400m Hurdles. by Kenny Guex. he women's 400m hurdles is a relatively

Linear profiling in the Oldenburg studbooks - from ideas to successful applications

Golfers in Colorado: The Role of Golf in Recreational and Tourism Lifestyles and Expenditures

Genetics Discussion for the American Black Hereford Association. David Greg Riley Texas A&M University

PUV Wave Directional Spectra How PUV Wave Analysis Works

Aerobic Capacity. Need Additional Resources?

Estimates of Genetic Parameters for Racing Times of Thoroughbred Horses

Life Transitions and Travel Behaviour Study. Job changes and home moves disrupt established commuting patterns

A REVIEW OF AGE ADJUSTMENT FOR MASTERS SWIMMERS

Influence of wind direction on noise emission and propagation from wind turbines

DANISH WARMBLOOD SOCIETY (DWB)

A Joint Nordic Animal Model for Milk Production Traits in Holsteins and Ayrshires

PEDIGREE ANALYSIS OF FOUR DECADES OF QUARTER HORSE BREEDING

To: ASHS Branches & Members

Stats 2002: Probabilities for Wins and Losses of Online Gambling

Price Determinants of Ranch Horses Sold at Auction in Texas. Authors

Surf Survey Summary Report

Progeny? Performance? What really matters in a stud fee. Charlotte R. Hansen and Sayed Saghaian. University of Kentucky

October 2010, Number 59 ALL-WAYS TM NEWSLETTER

There are 3 sections to the home page shown in the first screenshot below. These are:

Selecting Beef Bulls

ENGINEERING ANALYSIS OF THOROUGHBRED RACING. Rubin Boxer. Revelation Software

Legendre et al Appendices and Supplements, p. 1

Stud Book Changes for Discussion WHITE PAPER

2018 OHIO FACT BOOK. A statistical guide to the Thoroughbred industry in Ohio. Prepared by The Jockey Club

Relationships of Hip Height Measurements with Growth and Carcass Traits of Crossbred and Angus Cattle

EBVS are LESSONS FROM THE ASBP / 2. EBVs of Bulls Entered in the ASBP have provided a reliable Prediction of the Performance of their Progeny

Average Runs per inning,

2018 MICHIGAN FACT BOOK

STANDARD SCORES AND THE NORMAL DISTRIBUTION

Estimates of Genetic Selection Differentials and Generation Intervals for Four Paths of Selection

Safety Assessment of Installing Traffic Signals at High-Speed Expressway Intersections

Simultaneous Genetic Evaluation of Sires and Cows for a Large Population of Dairy Cattle

2018 MANITOBA FACT BOOK

2018 NEW JERSEY FACT BOOK

Gizachew Tiruneh, Ph. D., Department of Political Science, University of Central Arkansas, Conway, Arkansas

Phenotypic Standard of. Australian Brahman Cattle. Compiled by the Australian Brahman Breeders Association Ltd VERSION 06

Analyses of the Scoring of Writing Essays For the Pennsylvania System of Student Assessment

Across-Herd Genetic Evaluation of Swine

July 2011, Number 62 ALL-WAYS TM NEWSLETTER

Examining the racing performance and longevity in the Hungarian Thoroughbred population

Horse Farm Management s Report Writer. User Guide Version 1.1.xx

Figure 39. Yearly Trend in Death Rates for Drowning: NSW, Year

2017 WEST VIRGINIA FACT BOOK

Interstallion - on the way to an international genetic evaluation of sport horses

2018 PUERTO RICO FACT BOOK

BEEFMASTER BREEDERS UNITED SPRING SIRE SUMMARY

Participation in Structured Swimming Lessons : 0-13 Years

Analysing the effective population size in the partially closed and fragmented breeding population of the Trakehner Horse breed

Become Expert at Something

TRIP GENERATION RATES FOR SOUTH AFRICAN GOLF CLUBS AND ESTATES

save percentages? (Name) (University)

Transcription:

GENETICS OF RACING PERFORMANCE IN THE AMERICAN QUARTER HORSE: I. DESCRIPTION OF THE DATA 1'2 S. T. Buttram 3, R. L. Willham 4, D. E. Wilson 4 and J. C. Heird s Iowa State University, Ames 511 and Colorado State University, Fort Collins 8523 ABSTRACT More than 1 million racing records of American Quarter Horses were used to provide a description of Quarter Horse racing data. The data presented five racing distances (21,32, 366, 42 and 796 m). Finish time was used to measure racing performance. Means and variances for finish time increased with length of the race, but the distributions were similar for the five distances. Each distribution was skewed to the right and more peaked than a normal distribution. Repeated records were an important source of information and their use should improve the accuracy of genetic evaluations when comparing horses and sires across races and tracks. There was a tendency for more older horses and geldings to be found in the longer races. Two-year-old horses raced almost exclusively against other 2-yr-olds, and most 3-yr-olds raced with horses their own age. An interaction was detected between sex and age that was interpreted to be the result of differential selection rates among the sexes. Further study of age and sex effects is needed to develop factors for adjusting Quarter Horse racing data for use in genetic evaluation. (Key Words: American Quarter ltorse, Genetic Analysis, Breeding, Selection, Racing Performance.) Introduction Quarter Horse races are characterized as short races run at top speed for a matter of seconds. Horses start from a starting gate and usually race the entire distance on a straightaway or, where races are longer than 42 m, on a track with one turn. Quarter Horse races range in length from 21 to 796 m. The racing distances reported in this study are 21 m (22 yards), 32 m (35 yards), 366 m (4 yards), 42 m (44 yards) and 796 m (87 1 Journal paper no. J-1282 of the Iowa Agric. and Home Econ. Exp. Sta., Ames. Project no. 155. ~The cooperation and financial assistance of the American Quarter Horse Association is gratefully acknowledged. 3Present address: DeKalb Swine Breeders, 31 Sycamore Road, DeKalb, IL 6155. *Dept. of Anim. Sci., Iowa State Univ., Ames 511. s Dept. of Anim. Sci., Colorado State Univ.. Fort Collins 8523. Received October 12. 1987. Accepted May 23, 1988. yards). Younger horses tend to run in the shorter races in preparation for the more prestigious and lucrative races that are run at the quarter-mile (42 m) distance from which the "Quarter Horse" received its name. The American Quarter Horse Association (AQHA) maintains individual performance records on all horses that have raced in officially sanctioned pari-mutuel and nonpari-mutuel Quarter Horse races in the U.S., Canada and Mexico. These records constitute a large data base that can be used by Quarter Horse breeders in the selection of horses for racing performance. Records for the top young horses and progeny groups are summarized periodically in terms of total number of placings or money won. Ojala and Van Vleck (1981) found that time traits were more heritable than traits based on placings or money won and would be more useful in evaluating sires for racing ability. The only measure of racing ability used in this study was finish time. Adjustment factors for important sources of variation are unknown, as are the heritability 2791 J. Anim. Sci. 66:2791-2799

2792 BUTTRAM ET AL. and repeatability for finish time. The purpose of this paper is to provide a description of Quarter Horse racing data to aid in the further analyses and interpretation of results necessary to carry out genetic evaluations of American Quarter Horses for racing performance. Materials and Methods Individual racing records of horses that raced from January 1, 1971, to September 1, 1986, were obtained from AQHA. Racing distances included 21, 32, 366, 42 and 796 m. Information available on each horse included the registration number of the horse, year of birth and sex. The sire and dam registration numbers and birth-year of each of the parents also were included. In many cases, multiple records were obtained for individual horses. Each record included the track and date of the race, a unique race designation, the handicap weight and finish time measured to hundredths of a second. Handicap weight was the total weight carried by the horse including the jockey, tack and any extra weight assigned by the race track secretary. The original data included all records of horses with finish times. Only the first horse to cross the finish line was timed in each race, and times for all other horses were calculated from a photograph by converting into seconds the distance between the first horse and every other horse. Most races were electronically timed, but hand-held timers were used where electronic timers were not available or did not work. The data were initially sorted into five sets based on racing distance. Data from the 21-, 32-, 366-, 42- and 796-m races made up sets one (DS1), two (DS2), three (DS3), four (DS4) and five (DSS), respectively. Data from each distance were considered separately and summary statistics were calculated for each of these five sets. The data were examined for normality and for the presence of extreme values. Summaries also were made by years, months and tracks. The number of races per horse and progeny per sire were summarized for each set. Horses with individual records were classified into one of three groups based on the total number of races run within each distance. The number of horses in each group was calculated to provide a description of repeated records. Sires were classified similarly, based on progeny number, so that the number of progeny per sire could be summarized. Frequency distributions were calculated for tracks at which individual horses and sire progeny groups raced to help determine whether comparisons could be made across tracks or if racing was localized within tracks. To quantify the effects of selection, means of first records were calculated for horses with different numbers of total records. The first recorded 32-m race was considered to be the first record of a horse. Averages of first records were computed from DS2 for horses that raced only once, two to five times, six to 1 times and more than 1 times. Horses born later than 1982 were deleted so that all horses were given ample opportunity to race at least 1 times. Means were calculated by sex to determine whether differential selection rates existed among sexes. Summaries were made by sex and age (calculated as the difference between year of the race and birth-year). Various combinations of sexes and ages within races also were observed. Plots of sex by age subclass means were used to check for the presence of interaction. Means and SD for handicap weight were calculated for each data set. Mean racing times at each weight also were calculated, and the relationship between handicap weight and average time was examined. Results and Discussion More than 1 million records were included in the five original data sets (Table 1). Two of the data sets (DS2 and DS3) were large and, together, contained over 82% of the total number of records. The 42- and 796-m races each accounted for about 8% of the data, and DS1 contained less than 2% of the records. Means and SD for time increased with length of the race. The linear regression of average time on distance (r 2 =.997) showed that there was little evidence that horses were beginning to tire at these distances. Indeed, the average speed (m/s) for the entire race increased as distance increased up to 42 m. This points out the importance of both a quick start and ability to sprint in determining the racing performance of a horse. Certainly there is less "holding back" and "jockeying" for position in Quarter

RACING PERFORMANCE IN THE QUARTER HORSE 2793 TABLE 1. DISTRIBUTIONAL PROPERTIES OF RACING TIME BY DISTANCE Distance, m Item 21 32 366 42 796 No. of observations 16,54 561,738 368,846 92,664 94,25 Mean, s 12.788 18.745 21.5 22.972 47.653 SD, s.48.55.579.643 1.287 Minimum, s 11.4 16.29 19.18 21.2 44.3 Maximum, s 15.7 24.89 26.73 27.77 56.3 Best time a, s 11.62 17.2 19.18 21.2 44.3 Skewness 1.5 1.21 1.17.81.49 Kurtosis 2.3 4.4 7.99 1.85 21.8 aworld record time recognized by the AQHA as of September 1, 1986. Horse racing than is found in the longer Thoroughbred and Standardbred races. There is also possibly less "pace of the race" effect than that described by Tolley et al. (1983) in Standardbred trotters and pacers. Except for the obvious differences in means and variances, all the data sets had similar distributions. For example, the official best time for each distance was about three SD below the mean in each data set. The percentage of observations greater than three SD above the mean of each data set ranged from.8% to 1.2%, and the percentage greater than four SD ranged from.2% to.4% for the five data sets. The distributions were similar in shape; when tested with the Kolomogorov D statistic (SAS, 1985), each deviated (P.1) from a normal distribution. In each instance the distribution was skewed to the right and appeared more peaked than a normal distribution (Figure 1). Some skewness was expected because of the physical limit of the ability of the horse to run infinitely fast. This limit, however, did not seem to be as important in causing skewness as the fact that there was essentially no limit on how slow a horse could run. As can be seen from the range in the data, times in the 32-m race were located as far as 11 SD above the mean. However, relatively few extreme values were found in the very narrow upper tail of the distribution, and it was assumed that these values were the result of horses that were affected by extraneous nongenetic factors. For example, horses that do not have a chance to place are sometimes allowed to "coast" through the last part of the race. Another possible cause of skewness was injury during the race. At some point, however, it becomes difficult to decide between horses that were affected by nongenetic factors and those that were genetically slow. Objective criteria should be used to edit extreme values from these data. Wilson et al. (1986), using a subset of the same data used in this study, reported values of skewness and kurtosis similar to the ones in Table 1 and found that a natural log transformation reduced both skewness and kurtosis. However, the product-moment and Spearman rank correlations between estimates of breeding values obtained from transformed and untransformed data were.9996 and.9997, respectively. When times were removed that were more than three SD greater than the respective means of the five distances, lack of normality was not considered to be a major problem. The number of Quarter Horse races increased steadily each year from 1971 to 1983 r -I i -4 ~ IIln,... 17~ w s II 3 2.~ 21 3 ;2 3 FIq~H 111/~ S Figure 1. Frequency distribution for finish time from 32-m racing records.

2794 BUTTRAM ET AL. TABLE 2. NUMBER AND PERCENTAGE (%) OF HORSES WITH VARIOUS NUMBERS OF RACING RECORDS BY DISTANCE Number of racing records Distance, m 1 2-9 >9 21 8,395 (72.8) 3,122 (21.7) 11 (.1) 32 33,85 (26.9) 78,927 (62.9) 12,827 (1.2) 366 29,555 (32.1) 54,417 (59.1) 8,713 (8.9) 42 19,592 (5.4) 18,42 (47.4) 858 (2.2) 796 6,252 (39.) 7,456 (46.5) 2,339 (14.6) and seemed to stabilize after 1983. Trends toward faster times were evident in the yearly means of the shorter distances, particularly during recent years. A further study of phenotypic, environmental and genetic trends is in progress. Races were distributed throughout the year with some concentration during the summer (May through September). No trends in monthly means were found. The number of horses per race with official finish times ranged from 1 to 12, with 1 being the most common in the three medium-length races and 8 the most common in DS1 and DS5. The average race sizes were 6.8, 8.2, 8.3, 8.1 and 7.8 horses per race in the 21-, 32-, 366-, 42- and 796-m races, respectively. Most of the race tracks represented accounted for less than 1% of the records for any given data set. But in each case, there were three to five tracks that each accounted for as much as 5% to 17% of the races at that distance. There did seem to be some specialization in that tracks that dominated racing in the medium- and long-distance races were not the same tracks with large numbers of 21-m races. Of the five tracks with the most races in each data set, about two-thirds had mean times faster than the overall data set, indicating either better track management, superior horses, or both. Within-track SD for these tracks were always smaller than overall SD. The average number of records per horse were 1.4, 4.5, 4., 2.4 and 5.9 for the 21-, 32-, 42- and 796-m races, respectively. Horses in DS1 had relatively few repeated records (Table 2), and 72.8% of them raced only once at that distance. About 6% of the horses found in the two larger data sets (DS2 and DS3) had from two to nine records, and a number of horses raced more than 1 times. Horses in 796-m races seemed to have raced more times than horses in other races. Almost 15% of the horses in DS5 had more than 1 records each, and as many as 19 horses raced 1 times or more at 796 m. The percentage of horses that raced at multiple tracks in each of the data sets was related to the percentage of horses with repeated records. Data sets with a higher percentage of horses with repeated records also had a higher percentage of horses that raced at more than one track. In DS1, 91.6% of the horses raced at only one track, whereas the percentages were 48.5% and 54.3% for the horses in DS2 and DS3, respectively. About 5% of the horses in the 32-, 366- and 796-m races raced TABLE 3. NUMBER AND PERCENTAGE (%) OF SIRES WITH VARIOUS NUMBERS OF PROGENY BY DISTANCE Number of progeny Distance, m 1 2-9 >9 21 2,375 (59.3) 1,421 (35.5) 27 (5.2) 32 7,21 (44.9) 6,272 (4.1) 2,338 (15.) 366 6,54 (47.4) 4,99 (39.1) 1,731 (13.5) 42 3,856 (51.7) 2,872 (38.5) 728 (9.8) 796 3,469 (59.2) 2,128 (36.3) 267 (4.5)

RACING P E R F O R M A N C E IN THE Q U A R T E R H O R S E 2795 at five or more tracks, but that percentage was considerably lower for either DS1 or DS4. A study of sire progeny groups (Table 3) indicated that a surprisingly high percentage (44.9% to 59.3%) of the sires represented in each data set had only one progeny with performance records. Means were 2.9, 8., 7.2, 5.2 and 2.7 progeny per sire for DS1 through DSS, respectively. A few sires had rather large numbers of progeny. For instance in DS2, 1% of the sires had more than 11 progeny and more than 575 progeny records. The average number of progeny records per sire was 4.1, 35.9, 28.9, 12.4 and 16. for DS1 through DS5. Again, the number of tracks at which the progeny of a sire raced increased as the number of progeny per sire increased in that particular data set. In DS2 through DS5, more than 5% of the sires had progeny that raced at more than one track, and from 17.7% to 29.1% of the sires had progeny records at five tracks or more. These findings indicate that racing performance of individual horses is not localized entirely within tracks and that repeated records can help to create comparisons between individual horses across races and tracks. The amount of information contributed by repeated records should increase the accuracy of a genetic evaluation of horses for racing performance. This is particularly evident if racing performance is lowly to moderately heritable. Use of all records seems to have more intuitive appeal than using only best time for each year, as suggested by Ojala and Van Vleck (1981). Best annual racing time is affected by a combination of favorable environmental effects that occur simultaneously (Ojala et al., 1987). Unless these effects are removed, best time does not provide an accurate measure of overall racing ability. Mean handicap weight was approximately 55 kg and was almost identical in all five data sets. Standard deviations were in the range of 1.1 to 1.3 kg. There was little relationship between handicap weight and average performance, except that horses that ran at 55.3 kg were remarkably faster than horses that carried either more or less weight (Figure 2). This phenomenon was seen in all the data sets except DS1. In the 42-m races, horses were unusually fast when they carried either 55.3 or 54.4 kg. Interestingly, 55.3 and 54.4 kg were the most common weights carried and were the weights at which many horses were trained. It is difficult to draw conclusions from this relationship, but possibly, either faster horses run at these weights more often than slower horses or Quarter Horses are extremely sensitive to weights that deviate from those at which they were trained. Another consideration is that the better jockeys tend to ride at these standard weights even when they are allowed to carry less weight (D. Vails, personal communication). 19.2-19.1 19..2! CO 18,9 18.8 1.4 ~"i~ / ~a 16.7 18.6 ~1. 18.5 51,7 i i l i i i i i i i l i i f 52.6 53.5 54.4 116.3 56,2 57.2,~,1 ~. i i 59.9 HANDICAP WEIGHT, KG Figure 2. Relationship between handicap weight and mean finish time at each weight from 32-m racing records. Data labels indicate the percentages of the data represented at given weights.

2796 BUTTRAM ET AL. Evidence of selection was found in that horses with only one record had slower average first records than horses with two or more records (Table 4). A trend for horses with more lifetime races to have higher first-record averages strongly suggests that better horses were selected to race more times than poorer horses. In addition, some evidence of differential selection rates among the sexes was found. The difference in average first performance of mares that raced only once and those that raced two to five times was similar to that of geldings (.27 vs.26 s). But when horses with only one record were compared with horses with more than 1 records, the difference in average initial performance was larger 'in geldings than in mares (.6 vs.51 s). Put simply, mares and geldings received similar selection pressure based on early records, but geldings were selected more heavily on later records than mares. Many Quarter Horse races are classified according to age, money won, number of placings, etc., with the objective that all horses within a race are of equal ability. Race classifications were not known in this study, but more than 5% of the races at each of the three shorter distances contained horses of the same age. Many of these were races for 2-yr-old horses that rarely raced against older homes. A number of races for only 3-yr-olds were found in DS2, DS3 and DS4, but 4% or more of the 3-yr-olds ran in races with mixed ages. No differentiation was evident for age groups greater than 3 yr old or for sex with respect to race classification. The most common racing age for Quarter Horses (Table 5) was 2 yr at the two shortest distances, 3 yr in the 366- and 42-m races and 4 yr at the longest distance. The percentage of 2-yr-old records in each data set decreased from 44.2% in DS1 to 14.4% in DS4, and only a negligible number of 2-yr-old horses raced in the 796-m race. In addition, about 2% of the records in DS5 were by horses that were more than 6 yr old. These observations illustrate the tendency for younger horses to race in the shorter races, as previously described. There also was a trend for more geldings to be found in the longer races (Table 6). The percentage of geldings increased from 32% of the records in DS1 to 44% in DS4, whereas the percentage of stallions remained constant and the percentage of records from mares decreased. Geldings represented 71% of the records in DS5. Stallions not useful as breeding stock are routinely castrated and raced as geldings, which contributes to the increased number of geldings. Mares were removed from the racing circuit at younger ages than stallions, and geldings continued to race longer than either mares or stallions. The same occurrence also was seen within data sets. One explanation is that mares and stallions were removed from racing because they were selected for breeding stock. Because horses are not used for breeding and racing simultaneously, the differential culling rates observed may be the result of TABLE 4. AVERAGES OF FIRST RECORDS OF HORSES WITH VARYING NUMBERS OF TOTAL RECORDS (32 M) Number of total records Item 1 2--5 6--1 >1 Stallion No. of observations 3,85 Mean, s 19.36 SD, s.691 Mare No. of observations 8,191 Mean, s 19.367 SD, s.681 Gelding No. of observations 4,347 Mean, s 19.388 SD, s.658 6,771 3,36 1,725 19.53 18.842 18.76.626.536.52 18,468 7,429 2,83 19.98 18.922 18.858.632.559.57 1,117 4,79 3,822 19.131 18.915 18.793.589.525.486

RACING PERFORMANCE IN THE QUARTER HORSE 2797 +, ~ B 5~,q. ~o 9 o o +, ~ +, ~ ~ ~ ~. +, ~. +, o. +~ r e Z I..-, 9. ~ ~ o. z ~ +i ~ +f ~ +i ~ +i ~ +i ~ +i,.5, ~4 Z o. r~ ~ u4,-.1 o. ~St~ ~1 v.o.o o.o.o.o.o.o I. o,.~ z~ ~a oz~

2798 BUTTRAM ET AL. Z [- o~ ~ E z >, z o,..1 ~ ~ +~ t~ ~ +, e,i +, 9 ~,...; ~eq..,,,., I'N 9 -i t~ Ox,,~- " v,-, e,i d ~ +, ~ +, rz~",..g ~ +~ N t~ O differential selection rates among the sexes at different ages. Means for age groups (Table 5) showed that racing performance improved from 2 to 4-yr of age in most cases and decreased after age 6. Evidence of selection bias was seen in DS4, in which a small number of the better young horses were chosen to race. Horses in this data set did not seem to improve from 2 to 4 yr of age. Selection over time may have caused the means of older age groups to be biased downward. Means for sexes (Table 6) indicated that geldings were slightly faster than stallions when summarized over all ages and that mares were slowest. Standard deviations for racing time also were higher in mares. Plots of sex by age subclass means (Figure 3) revealed an important interaction that was seen in each of the four shorter races. As 2-yr-olds, geldings ranked between mares and stallions for performance in the 21-, 32- and 366-m races and slower than either of the sexes in DS4. However, by 4 or 5 yr of age, geldings were faster than either mares or stallions and continued to rank fastest as age increased. No interaction of gender and age was found in DS5 because geldings were fastest at all ages (2-yrolds not included). The possibility of a biological explanation for the interaction cannot be overlooked, but further study (Buttram et al., 1988) showed that the effects of age for stallions and geldings were very similar. The interaction was interpreted to be the result of differential selection rates among the sexes. As age increased, the slower geldings were removed from the racing circuit, leaving only faster geldings to contribute to the means of older age groups. Conversely, faster stallions and mares were removed at relatively young ages to be used for breeding stock. The difference in racing performance between stallions and mares is illustrated by the lines in Figure 3. The lines for mares and stallions were almost parallel from ages two to four in each of the data sets. Some spread in the difference between the two sexes occurred after 4 yr of age. From these observations it is not clear how much effect selection had on age and sex means. Any bias caused by selection should be removed to increase the accuracy of estimating effects of sex and age. Sex and age effects seem to be important and should he included when genetically evaluating Quarter Horses for racing performance, but further study is needed to

RACING PERFORMANCE IN THE QUARTER HORSE 2799 19. 18.9 9 STAL.UON8 4 IANRE8 18.8 18.7 i 18.6 / 18.5 J 18.4 r i t i i i i 2 3 4 5 7 8 AGE., YR Figure 3. Relationship between age and mean finish time at each age by sex from 32-m racing records. develop appropriate adjustments for these effects. L iteratu re C ited Buttram, S. T., R. L. Willham and D. E. Wilson. 1988. Genetics of racing performance in the American Quarter Horse: I1. Adjustment factors and contemporary groups. J. Anim. Sci. 66:28. Ojala, M. J. and L. D. Van Vleck. 1981. Measures of racetrack performance with regard to breeding evaluation of trotters. J. Anim. Sci. 53:611. Ojala, M. J., L. D. Van Vleck and R. L. Quaas. 1987. Factors influencing best annual racing time in Finnish horses. J. Anim. Sci. 64:19. SAS. 1985. SAS Users Guide: Basics. SAS Inst., Inc., Cary, NC. Tolley, E. A., D. R. Notter and T. J. Marlowe. 1983. Heritability and repeatability of speed for 2- and 3-yr-old Standardbred racehorses. J. Anita. Sci. 56: 1294. Wilson, D. E., R. L. Willham, S. T. Buttram and J. C. Heird. 1986. Genetic evaluation of American Quarter Horses for racing time using a reduced animal model and accounting for repeated records. J. Anita. Sci. 63 (Suppl. 1): 18. (Abstr.).