Feature Race | Auction | Breeding | General | Hall of Fame | Harness | Interviews | Mixed Breed | New Players | Racing | Site Updates | Steeplechasing | Steward's Cup | Triple Crown
Revisiting Workout Times vs Speed Figures
Original article written by Wc Lee posted 2 months 2 weeks agoApproximately a year ago, I wrote a series of articles comparing Workout Times to Speed Figures. The goal of the articles was to see if we could use Workout Times to predict how well horses would run in a SIM race. Those articles no longer exist, as the SIM has a policy of deleting articles that are older than 6 months. The reason for the auto deletion is to prevent the SIM from being overloaded with newbie how to articles, race summaries and other topics that did not have long term value.
In hindsight, I am glad that my articles got deleted, as the results of the analysis were flawed. The methodology was not flawed, but the underlying assumptions that went into the analysis was incorrect. When I wrote the articles, I had been playing the SIM for about 6 months. The SIM is deceptively simple, but there are nuances that took me a long term to figure out.
I made three incorrect assumptions in my analysis:
1) All Speed Figures are created equal.
Speed Figures are the primary means for players to judge how well a horse performs. However, the Speed Figure is very dependent upon the level of competition. A horse that runs a Speed Figure of 80 in a Maiden race is not as good as a horse that runs a Speed Figure of 80 in a Grade 1 race. If you try to run a horse that ran an 80 Speed Figure in a Maiden race in a Graded race, I will guarantee you that the horse will run a Speed Figure much worse than 80. In my original analysis, I used the Speed Figures from all of the races that a horse ran in the first 8 weeks that that horse is eligible to run. So I mixed Speed Figures from Graded Races, Allowance Races, Claiming races and Maiden races.
2) The Speed Figure for a horse is representative of a horse’s ability no matter where the horse finishes in a race
In my original analysis, I used the Speed Figures for all of the races that a horse ran in the first 8 weeks of eligibility. Even if the horse ran last in a 12th field race, I used the Speed Figure. What I have found since then is that once you get past the 3rd finish position, the Speed Figure that a horse runs is no longer representative. In this way, the SIM simulates real life. When you watch a real life race, you will see that the jockeys in horses that in the back of the pack will ease up and stand up if the horse is not going to finish in the money. As a result, I should have disregarded the Speed Figures for those horses that did not finish in the top three positions.
3) The fastest Workout times for a given set of equipment and distance is the most representative of a horse’s ability
In my original analysis, I used the best workout time for each workout distance and equipment. I have since come to realize that the SIM adds/subtracts a random time to each workout, so the best workout time is not representative. I should have taken the average workout time if a horse ran multiple workouts with the same equipment for a certain distance.
Recently there was a forum post that asked where you could find my original articles. I never kept archived copies of any of the articles that I wrote for the SIM, so when the SIM deleted the articles, the articles were gone for good.
After I read the forum post, I thought this would be an opportune time to for me to update my analysis and to correct the mistakes that I had done earlier.
In this updated analysis, I compared the Workout Times to Speed Figures for approximately 18,500 Thoroughbred horses that have recently raced in the SIM. The total number of horses that I looked at was over 150,000, but I eliminated horses using the following factors.
1) I only use Speed Figures from Maiden Special Weight races. If a horse ran only in Maiden Claiming or Claiming races, the horse was not included in the analysis. Maiden Claiming and Claiming races have easier competition than the Maiden Special Weight races; in order to do apples to apples comparisons, I only used Speed Figures from the regular Maiden Special Weight races.
2) I only use Speed Figures for horses that finish in the top 3 of a Maiden Special Weight race. This eliminated many horses, especially in the Dirt Router division, where the 2 Year Old races often have very large fields.
3) I only consider 2 Year Old Maiden Special Weight races during Weeks 1 thru 8 for Sprinters and only 2 Year Old Maiden Special Weight races for Milers and Routers. So if a horse did not run in a race until it was 3 years old, it was eliminated from the analysis. This avoids having to factor in gallop changes in the analysis.
4) If a horse did multiple workouts with the same equipment for a certain distance, I used the average of the workout times.
5) If a horse ran a race with a set of equipment and did not run a workout with the same set of equipment, the horse was eliminated from the analysis.
For those of you who want to see the data, I saved an Excel spreadsheet on my Google Cloud drive in this location:
https://docs.google.com/spreadsheets/d/1XWTNyFBeqsLW88sY--LmkT8Ys4rXTubG/edit?usp=sharing&ouid=111904769154589212428&rtpof=true&sd=true
Dirt Sprinters
----------------
There are 2 workout distances for Sprinters, 2 Furlongs and 3 Furlongs. There were approximately twice the number of horses (2430 vs 1234) that ran workouts at 3 Furlongs than ran workouts at 2 Furlongs. The correlation coefficient for the horses that ran at 2 Furlongs was slightly better (-0.96) than was the correlation coefficient for the horses that ran at 3 Furlongs (-0.82). Both correlation coefficients showed a strong linear relationship between Workout Times and Speed Figures.
This is the regression equation for 2 Furlongs is:
Speed Figure = 707.262 - Workout Time x 28.767
This is the regression equation for 3 Furlongs is:
Speed Figure = 789.377 - Workout Time x 20.155
Dirt Milers
------------
The standard workout distance for Dirt Milers is 4 Furlongs. There were 1775 horses that met the analysis criteria and the correlation coefficient was -0.79. This shows a show strong linear relationship between Workout Times and Speed Figures.
The regression equation for 4 Furlongs is:
Speed Figure = 777.151 - Workout Time x 15.286
Dirt Routers
--------------
There are 3 workout distances for Dirt Routers: 5 Furlongs, 6 Furlongs, and 7 Furlongs. By far, the predominate workout distance is 5 Furlongs (2658 horses) with 6 Furlongs (647 horses) and 7 Furlongs (46 horses) trailing far behind.
The Correlation Coefficient for the Dirt Router workouts is not as good as the Correlation Coefficients were for the Dirt Sprinters and Dirt Milers. The Correlation Coefficients were for 5 Furlongs -0.59, for 6 Furlongs -0.63 and for 7 Furlongs -0.43.
There are too few horses for the 7 Furlong workout distance for the Correlation Coefficient and the corresponding regression equation to be meaningful.
The Correlation Coefficient for both 5 Furlongs and 6 Furlongs show a moderate linear relationship between workout times and Speed Figures. The cutoff between moderate and strong in terms of Correlation Coefficients is +/- 0.7.
I believe there are two reasons why the Correlation Coefficient is lower for the Dirt Routers.
First, there are a wide variety of race distances that for a Dirt Router to run (from 1 1/16 miles to 2 ½ miles). A horse that has its best performance at the longer distances may not do as well in the shorter router distances. However, in the vast majority of cases, the distance that a 2 Year old Dirt Router can run in is at 1 1/16 miles.
Second, as mentioned earlier, the field sizes for Dirt Routers is very large. In an article that I published last year, I showed that the average field size for Dirt Router races in the SIM was almost 2 horses larger than the average field size for most other Thoroughbred divisions. Large field sizes will depress Speed Figures for horses that finish out of the money. If you have a Stakes horses, if there are 3 other Stakes or better horses in the field, the Speed Figure will not be representative of ability for the Stake horse that did not finish in the top 3. Even the 2nd and 3rd place horses' Speed Figures may not be representative if the winner was far out in front.
The regression equation for 5 Furlongs is:
Speed Figure = 824.287 - Workout Time x 12.874
The regression equation for 6 Furlongs is:
Speed Figure = 1300.191 - Workout Time x 17.485
Turf Sprinters
---------------
As with the Dirt Sprinters, there were almost twice as many horses that did workouts at 3 Furlongs (2352 horses) as there were horses that did workouts at 2 Furlongs (1473 horses). The Correlation Coefficients for 2 Furlongs (-0.82) and 3 Furlongs (-0.78) both showed a strong linear relationship between Workout Times and Speed Figures.
The regression equation for 2 Furlongs is:
Speed Figure = 963.790 - Workout Time x 39.087
The regression equation for 3 Furlongs is:
Speed Figure = 946.942 - Workout Time x 24.086
Turf Milers
-----------
There were 1754 horses that ran a Maiden Special Weight race with the same equipment as they did when running a 4 Furlong workout. The Correlation Coefficient of -0.88 showed a strong linear relationship between Workout Times and Speed Figures.
The regression equation for 4 Furlongs is:
Speed Figure = 775.892 - Workout Time x 14.924
Turf Routers
-------------
As with Dirt Routers, the predominant workout distance was 5 Furlongs (2041 horses), with 6 Furlongs (516 horses) and 7 Furlongs (21 horses) far behind. However, unlike with Dirt Routers, the Correlation Coefficients for 5 Furlongs (-0.71) and 6 Furlongs (-0.71) showed a strong linear relationship between workout times and Speed Figures. I disregarded the results for 7 Furlongs (Correlation Coefficient of -0.60) because of the low number of horses that ran this workout distance with the same equipment as they did running a Maiden Special Weight race.
I attribute the higher Correlation Coefficient for the Turf Router case vs. the Dirt Router case to the smaller field size of Turf Router races.
The regression equation for 5 Furlongs is:
Speed Figure = 1064.998 - Workout Time x 16.836
The regression equation for 6 Furlongs is:
Speed Figure = 1763.296 - Workout Time x 24.005
All Weather Sprinters
-----------------------
There are far fewer All Weather horses than there are Dirt horses or Turf horses in the SIM. For the All Weather Sprinters, there were 260 horses than ran a workout distance of 2 Furlongs with the same equipment as a Maiden Special Weight race. There were 408 horses that ran a workout distance of 3 Furlongs.
As with the Dirt Sprinters and the Turf Sprinters, the Correlation Coefficients for both the 2 Furlongs (-0.72) and 3 Furlongs (-0.77) showed that there is a strong linear relationship between Workout Times and Speed Figures.
The regression equation for 2 Furlongs is:
Speed Figure = 731.305 - Workout Time x 27.385
The regression equation for 3 Furlongs is:
Speed Figure = 892.482 - Workout Time x 21.860
All Weather Milers
--------------------
There were 343 horses in the study that ran an All Weather 1 mile race with the same equipment as they did for a 4 Furlong workout. The Correlation Coefficient of -0.62 showed a moderate linear relationship between Workout time and Speed Figure.
I attribute the smaller Correlation Coefficient to the lower variability of horse ability of All Weather Milers. There are fewer All Weather Milers than any other Thoroughbred Division in the SIM. The separation of Workout Times for middle 80% of All Weather Milers that finished in the Top 3 in a Maiden Special Weight race was just 0.29 seconds (47.50 seconds to 47.79 seconds). In contrast, the separation of Workout Times for the middle 80% of Dirt Milers was almost three times greater at 0.76 seconds (45.27 seconds to 46.03 seconds).
The regression equation for 4 Furlongs is:
Speed Figure = 700.015 - Workout Time x 13.020
All Weather Routers
----------------------
There were 490 horses that ran an All Weather Router Maiden Special Weight race with the same equipment as they did a 5 Furlong workout. There were 114 horses that ran All Weather Router race and did an 6 Furlong workout and just 5 horses that did the same for a 7 Furlong workout.
The Correlation Coefficients for the 3 workouts were -0.67 for 5 Furlongs, -0.75 for 6 Furlongs and -0.37 for 7 Furlongs.
I think that there are too few horses to use the regression equations for 6 Furlongs and 7 Furlongs.
The regression equation for 5 Furlongs is:
Speed Figure = 1085.304 - Workout Time x 16.862
How to Use the Regression Equations
------------------------------------------
I use the conversion equations to get an idea as to how “good” my horses are before they run in an actual race. I find that the Gallop of a horse sometimes does not translate to actual performance. On the other hand, I find that workout times are a pretty good indicator of a horse's current ability.
I also use the conversion equations to determine if it is worthwhile to enter a horse into a Maiden Special Weight race, especially during the first few weeks that horses are eligible to race, when there are no Speed Figures available.
Finally, I use the conversion equations to determine how “good” another player’s horse is when it is entered in a race with one of my horses that has already raced.