Do Speed Figures Increase With Distance?

Writer’s Note: If you just want to get to the results and don’t care about the background, data source,  methodology, or analysis, you can scroll down to the bottom (and maybe check out the pretty pictures on the way).

Introduction

For those of you not familiar with real life horse racing (probably very few of you), there is something called a Beyer Speed Figure assigned to every horse and race in the US.  The Beyer Speed Figure is “a numerical expression of a horse’s final time, universalized for distance, track surface, and the daily variant on that surface.”  The idea behind it is pretty simple.  A horse finishing 10 furlongs on the Churchill Downs dirt on the first Saturday in May in 2:02 may have actually run faster, or slower, that an equivalent horse running 10 furlongs on the Belmont Park dirt on the first Saturday in May in 2:02.  And it may be slower or faster that a horse finishing in 2:02 on a different day.  Or on the turf.  And using the 2:02 time wouldn’t be very helpful if you want to compare how a horse did at 10 furlongs versus how it did at 6 furlongs, nor would the 6 furlong split time in a 10 furlong race be helpful (horses need to carry their speed for longer, so the opening fractions will be slower at longer distances).  So the idea behind its creation was to assign a single number, that can be used across tracks, distances, and dates.  A 120 Beyer Speed Figure indicates a race was “faster” than a 100 Beyer Speed Figure, regardless of the time of the race.

The Sim has a similar stat referred to as the Speed Figure (or “SP”, as it will be referred to throughout the rest of this article).  The Sim Speed Figure is not quite the same thing as the Beyer Speed Figure (it does not change based on track variant), but the idea is the same – It is a way to compare horses across distances, tracks, and surfaces.  In theory, a 120 SP in the Sim at 8 furlongs on the turf is “faster” than a 105 SP  at 10 furlongs on the dirt, even if that 120 SP ran second in a NW2x allowance and the 105 SP won a stakes race.  It does NOT mean that the horse that ran the 120 SP would beat the horse than ran the 105 SP if they had raced against each other, because some horses prefer certain distances, surfaces, and possibly even tracks.  But Speed Figures in the Sim are a good way to determine if your horse likes a particular surface, distance, or even running style.

But in order for the above to work to allow you to use Sim Speed Figures to determine your horse’s best distance, Speed Figures need to be consistent across all distances.  In other words, if your horse ran a 120 SP at 8 furlongs, then if he ran the EXACT SAME race at 6 furlongs, the Speed Figure should also be a 120.  We can assume this is true, but without seeing the actual code that makes up the game, we can’t know for sure.

What we DO know, however, is the actual Speed Figures that every horse has earned in every race in the Sim.  This gap in knowledge led to a few questions and anecdotal observations posted on the Sim Message Board, which are:

1)      Dkyoung: “I was wondering why Speed Ratings seem to increase with distance as a rule.  Shouldn’t the top speed ratings be consistent no matter what distance raced? … For example, if a horse rated 100 over 4.5f then rated 100 over 8f next start it is clearly not the same level of performance in the sim.”

2)      Dixiedotco: “Don’t know about the longer distances, but this is certainly true when it comes to the sprint distances. A horse that runs, for example, 90 at 4.5f in a winning effort will probably run somewhere closer to 100-105 going 7.5f….7f and 7.5f races especially lend themselves to large speed figures that may not accurately represent the horse’s ability when compared to other horses that run shorter.”

3)      Tyedye: “There’s always been a long time discrepancy (inflated speed figures) for races at 7.5f and 10.5f. I think it’s some sort of rounding error in however make comes up with the speeds for these distances.”

4)      Itch: “Shaggy your stuff should be in a BTB article.”

Well, without knowing the Sim’s code, we don’t have all the answers.  But we do have data.  And we can perform some analysis of that data to try and answer these questions (well, except for #4…that one is answered by this article being published).  We will boil those comments down to these main questions:

1)      Do Speed Figures increase with distance?

2)      Do races at 7.5 furlongs and 10.5 furlongs have inflated Speed Figures?

3)      Is a Speed Figure at one particular distance “equivalent” to a Speed Figure at a different distance?

4)      Can I write a BTB article analyzing the data that anyone will read?

So with that (long) introduction, it’s time to play one of my favorite Sim side-games: FUN WITH NUMBERS!

Methodology

The main thing to understand before we get into the analysis is what data is being used.  The analysis was performed on a set of data comprised of all 2017 Races.  There are multiple reasons for limiting the data in this way.  One is just practicality: the Sim has been in existence for 19 years, and if I were to go back to 1998, that would be way too much data for my computer to handle.  The second is for consistency of data: The Sim has gone through many changes over those 19 years.  The calculation of Speed Figures was changed back in 2004, so anything prior to that would have needed to be converted into present-day Speed Figures (similar to how you hear “X is the highest grossing movie of all-time…adjusted for inflation).  Furthermore, there has been a SIGNIFICANT increase in Speed Figures over the past 2 years.  How significant?  From 2005-2015, there were 66 Speed Figures over 120 in the Sim.  In 2017 ALONE, there were 103 Speed Figures over 120 in the Sim. We had 56% more Speed Figures over 120 in 2017 ALONE than we had for an 11-year stretch from 2005-2015 COMBINED.  This is a sign of a different problem in the Sim, but that is subject for a different analytical deep dive.  For purposes of this article, the important thing is that because of this drastic difference, I have limited Speed Figure data to 2017 only.

But limiting data to 2017 alone isn’t enough.  2 year olds are always slower than their older counterparts.  3 year olds are typically slower early in the year than their older counterparts.  Different surfaces have different speed figures.  So from the full set of races from 2017, I made the following changes:

(1) all 2yo races were removed

(2) all 3yo races prior to July 1, 2017 were removed

(3) all Polytrack races were removed (there are far fewer of them, as the CPU does not card polytrack races)

(4) All off-track races were removed

(5) Turf and Dirt races were analyzed separately

After those changes, there were 322,141 Dirt Races and 312,214 Turf Races in 2017.  But just making those changes still leaves some concerns:

(1) Female horses are typically slower than their male counterparts

(2) There is no control for wear and tear and the number of races a horse had previously run (a horse in its 50^th race will be slower than a horse in its 20^th race)

(3) There is no control for horses running at non-ideal distances/surfaces (for example, a sprinter trying a 16 furlong race)

(4) There is no control for injuries or effort (A 1.2 SP due to injury is not indicative of the horse’s ability or speed figures generally at that distance).

To minimize the impact that those limitations have on the analysis itself, I then used five different methods for analyzing the data:

(1) The AVERAGE Speed Figure at each distance (referred to as “AvgSP”), which is an average of each speed figure at the distance, regardless of how high or low.

(2) The MEDIAN Speed Figure at each distance (referred to as “MedianSP”), which is the median speed figure at each distance.  This minimizes the impact of outliers on the data (for example, if there were 5 races in total and the Speed Figures were 100, 100, 100, 100, and 2, the MedianSP would be 100, but the AvgSP would only be 80.4).

(3) The MAXIMUM Speed Figure at each distance (referred to as “MaxSP”), which is the fastest speed figure by any horse at the distance.

(4) The AVERAGE OF THE TOP 100 Speed Figures at each distance (referred to as “100SP”).  This looked at the fastest 100 Speed Figures at each distance and took the average of those (the concern about outliers is negated by the fact that Speed Figures that would normally drag down the average would not have fallen within the top 100).

(5) The AVERAGE OF ALL SPEED FIGURES OVER 70 SP at each distance (referred to as “70SP”).  This has two benefits.  First, it attempts to remove injuries or wear and tear from the overall analysis, as horses who perform poorly are dropped out without eliminating a majority of the races at the distance).  Second, it allows us to see just how many races there are at each distance where the Speed Figure is over 70 (based on the initial questions, one might suspect that a 7f race will have a higher percentage of speed figures over 70 than a 4.5f race).

There was one final alteration made to the data that was not initially part of the plan, but was somewhat necessary once the analysis began.  Certain distances (11.5f, 12.5f, 13.5f, 14.5f, and 15.5f)  are not carded by the CPU, and therefore the number of races at the distance is very low.  A low number of races increases the likelihood of error in the data and makes it harder to detect trends.  Therefore, I determined that I would only analyze data for distances that made up at least 1% of the races carded during the year.  As a result, races at 11.5f, 12.5, 13.5, 14.5f, 15f, and 15.5f were removed from the trend analysis (15f on turf was the toughest decision, as 0.99% of turf races in 2017 were run at 15 furlongs, but ultimately the average SP (62.39449) was so much lower than any other distance that it made more sense, even if just for graphical purposes, to eliminate it from the data.  I’m also not sure if 15f races are carded by the CPU; they were not in the Old Sim).

Results and Analysis

Overall, we can see similar trends in both the dirt and turf races that answer our three initial questions.  To answer the first question posed above (whether Speed Figures increase over distances), Speed Figures in the Sim DO increase as distances increase, but only up to a point and not uniformly.  Speed figures increase exponentially from 4.5 furlongs to 8.25 furlongs, then decrease exponentially (though with a slower decline) from 8.25 furlongs to 16 furlongs, leveling out as we hit the long distances.

The second question posed above (whether 7.5 furlong and 10.5 furlong races have abnormally high speed figures) is significantly unlikely.  In fact, based on the data used for this study, 7.5 furlong and 10.5 furlong races seem to have LOWER speed figures than would be expected.

For the third question posed above (whether speed figures are equivalent across distances), the data does not give a definitive answer, but we can draw some conclusions.  There is every indication that horses running at short distances (4.5 furlongs) will put up slower speed figures than those running at slightly longer distances in sprints, (7.5 furlongs), and that the opposite is true for routes.  Overall, it appears that a top level horse will need to put up a much higher Speed Figure when running in the 6 furlong to 10 furlong range than it will need to at either end of the spectrum (shorter than 6f or longer than 10f).

Now it’s time to look at some of the data:

AvgSP:

Fig. 1 – Dirt AvgSP and Turf AvgSP data.  Bubble size indicates the proportional number of races at each distance across all dirt/turf races (the larger the bubble, the more races).  Red line indicates the average SP for the data across all distances (points above the red line are distances that have an AvgSP higher than the overall average; points below the red line are distances that have an AvgSP lower than the overall average). 

The above charts show the AvgSP on both dirt and turf for all of 2017.  The size of the blue bubbles indicates the number of races at that distance – the bigger the bubble, the more races were run at that distance.  The red line in both charts shows the AvgSP of all dirt/turf races from the year (the average dirt SP was 84.54459, the average turf SP was 83.60584).  The black line in each is the trendline for the data, based on a cubic exponential equation (which is also listed in the charts).  The important number to look at is the R² value.  This shows you the percentage of data that is explained by the trendline.  An R² value of 0 would essentially mean that the trendline is useless, whereas an R² value of 1 would mean the trendline perfectly explains all of the data.

If Speed Figures were the same across all distances, we would expect that the average at each distance would be the same and thus the data would form a straight line (similar to the red line above).  But that is not the case; rather, the data seems to rise and fall.  Furthermore, if there was a consistency to the increase or decrease (which would happen if, say, speed figures increase as distance increases), we would expect that a linear trendline would explain a lot of the data.  But in this case, it does not.  A sample of that is below, where the linear trendline for turf races from just 4.5f to 8f has an R² value of 0.9766 and the linear trendline for turf races from 8f to 16f has an R² value on 0.8581 (both very close to 1):

Fig. 2 – Turf AvgSP separated into sprint races (left) and route races (right).  Bubble size indicates the proportional number of races at that distance across all turf races. 

But that is not the case for the full set of data.  The linear trendline for all turf races has an R² value of only 0.138, indicating that a linear trendline does not explain very much of the overall data.  Instead, an exponential trendline fits the data far better (specifically a 3^rd power exponential trendline), which indicates that Speed Figures rise to a certain point, then drop to a certain point before flattening out.  As you can see in the initial charts above, the R² value for Dirt AvgSP is 0.915, and the R² value for the Turf AvgSP is 0.8828, indicating that the trendline explains approximately 91.5% and 88.28%, respectively, of all of the data.

In general, at sprint distances, the AvgSP increases by approximately 2 for every additional half furlong, up to 8 furlongs.  For route races, the AvgSP drops roughly 1.5 for every additional half furlong, beginning at 8 furlongs and going up to 16 furlongs.  Based on those statistics, if a horse put up a 76 SP in a 5 furlong race, the same race would have earned that horse an 88 SP in an 8 furlong race, while that same race at 16 furlongs would have earned a 64 SP.

One interesting data point to look at is the difference between Speed Figures at 12 furlongs on the turf and on the dirt.  Most data points, whether on dirt or turf, are in the same general vicinity away from the trend line.  But 12 furlongs has a wide disparity between surfaces.  On dirt, the AvgSP of a 12 furlong race is only slightly above where the trendline would expect it to be (the trendline expects a 12 furlong AvgSP on the dirt to be 77.2766, whereas the actual AvgSP is 78.97722, approximately 1.7 points higher than the trendline).  On turf, however, the AvgSP of a 12 furlong race is above the baseline, and FAR above the trendline (the trendline expects a 12 furlong AvgSP on the turf to be 79.1373, whereas the actual AvgSP is 84.00089, approximately 4.9 points higher than the trendline).  This is possibly due to two different causes: one is that there are significantly more 12 furlong turf races than dirt races (10,951 (or 3.51% of the total) on turf, versus 6,859 (or 2.13% of the total)); the other is that the class of horses running at 12 furlongs on the turf is higher than the number running at 12 furlongs on the dirt, due to the “classic” turf distance being 12 furlongs on the turf vs. 10 furlongs on the dirt (there were 84 Graded Stakes at 12 furlongs on the turf in 2017, versus only 8 on the dirt).  What this means is that the turf data at 12 furlongs might be more accurate than the dirt data

MedianSP:

Fig. 3 – Dirt MedianSP and Turf MedianSP data.  Bubble size indicates the proportional number of races at each distance across all dirt/turf races (the larger the bubble, the more races).  Red line indicates the median SP for the data across all distances (points above the red line are distances that have a MedianSP higher than the overall median; points below the red line are distances that have a MedianSP lower than the overall median). 

Doing a calculation based on MedianSP had the benefit of minimizing the impact of outliers on the data.  But ultimately, the trend analysis is relatively similar to that of AvgSP.  The R² value for dirt MedianSP and turf MedianSP is 0.899 and 0.8746, respectively, within two percentage points of the data from AvgSP.  The data shows that the MedianSP increases as distances go from 4.5 furlongs to 8.5 furlongs, then decrease from 8.5 furlongs to 16 furlongs, leveling off towards the end.

The main difference here is the overall increase in Speed Figures at each distance.  The MedianSP is anywhere from 2-5 points higher at each distance than the AvgSP at the distance.  This isn’t surprising, given that the outliers we were trying to correct for were injuries and extreme wear and tear; all of those outliers would be on the low end of the SP spectrum.  As you can see, the baseline MedianSP is an 87.87622/86.82092 (Dirt/Turf), which is approximately 3.3 points higher than the baseline AvgSP of 84.54459/83.60584 (Dirt/Turf).  The difference in MedianSP from AvgSP gets higher as the distances increase.  For sprint races (4.5f-7.5f), the difference between MedianSP and AvgSP is only an average of 2.6/2.3 (Dirt/Turf); for standard route races (8f-10f), the difference between MedianSP and AvgSP is 3.5/3.4 (Dirt/Turf); for long route races (10f-16f, minus the excluded data), the difference between MedianSP and AvgSP is 4.2/4.6 (Dirt/Turf).  This indicates that the outlier data more severe the longer the races get.  Again, this is unsurprising – Every horse can run 4.5 furlongs; very few have the stamina to go a full 16 furlongs.

Fig. 4 – Dirt MedianSP separated into sprint races (left) and route races (right).  Bubble size indicates the proportional number of races at that distance across all turf races. 

We can also see, from the above charts of MedianSP on the dirt, a similar linear correlation when splitting up distances into sprints (4.5 furlongs to 8 furlongs) and routes (8 furlongs to 16 furlongs).  The data is again very similar to the AvgSP data, shifted upward for increased MedianSP across every distance.  In fact, the linear relationship in general is the same as AvgSP – the MedianSP increases approximately 2 points for every half furlong increase in sprints, and decreases by approximately 1.5 points for every half furlong increase in routes.

MaxSP:

Fig. 5 – Dirt MaxSP and Turf MaxSP data.  Data represents the fastest SP at each distance during the data set. 

Once again, we have a similar trend when looking at the MaxSP for each distance (you’ll note that this is a common theme of this article).  MaxSP tends to rise as you increase the distance in sprint races, and decline as you increase the distance in route races.  There are, however, a couple of unique items to look at here.  First, on the dirt side, the MaxSP increases from 4.5 furlongs to 8.5 furlongs and then decreases, but actually rebounds and increases slightly at 16 furlongs (data from AvgSP and MedianSP merely flattens out more at the 16 furlong distance).  One possible culprit for this is that the 16 furlong data is an outlier – in other words, someone ran a REALLY good race.  While MaxSP removes the possibility of any injury or wear and tear, it has a high sensitivity to outliers on the upper end of the spectrum.  On December 24, Golfteam, a 4yo gelding owned by golfteam1, took home G1 The Hit Man in Italy by 4.25 lengths, with a speed figure of 115.2099.  The next fastest SP for the year was the second place finisher, New King In Town, owned by frankster, who put up a 112.7563 SP.  The next fastest was a 111.65918 SP.  If you remove the 115 outlier, the trend line for dirt MaxSP looks very similar to the turf MaxSP – rising to the middle distances, then falling for the longer distances.

The turf MaxSP chart is more standard in that the trendline rises and then falls.  Where that gets interesting is at the pivot point – Most other charts show a decline in Speed Figures beginning around 8 or 8.5 furlongs.  The trendline for turf MaxSP, however, keeps increasing until it reaches 9.5 furlongs.  So why is the pivot point of the trendline at such a longer distance than the others?  I don’t have a good explanation for it.  There’s no obvious answer like switching from sprints to routes or having a main distance for the surface (like 10 furlongs on the dirt or 12 furlongs on the turf).  So it’s an interesting fact to find.

Even more interesting is that the actual data for 9.5 furlongs is significantly below the trendline.  The actual MaxSP of 118.937 is approximately 3 points below the trendline expectation of 121.875, the biggest difference in actual data vs. trendline data.  The only other data in the turf MaxSP analysis close to that is the 8.25 furlong distance, which is an inherently flawed distance in the MaxSP analysis because there were 0 stakes races run at the distance in 2017, thereby limiting the maximum Speed Figure that could be achieved at the distance.  If we were to remove the data for both 8.25 furlongs and 9.5 furlongs, the R² value of the turf MaxSP trendline would jump from 0.8401 to 0.9222.  Again, there does not seem to be an obvious reason for this.  Yes, there are fewer graded stakes at the 9.5 furlong distance (8) than the surrounding distances (9 furlongs had 71 graded stakes, 10 furlongs had 93 graded stakes), so there is less potential for a high SP.  But there were plenty of regular stakes that could have seen a higher SP (there were 163 stakes races at 9.5 furlongs on the turf in total).

100SP:

Fig. 6 – Dirt 100SP and Turf 100SP data.  Red line indicates the average SP of the full data set comprised of 100 fastest SPs at each distance (points above the red line are distances that have a 100SP higher than the overall average; points below the red line are distances that have a 100SP lower than the overall average). 

There isn’t much to say about the 100SP data that hasn’t already been said above.  The 100SP list limits all of the data to only the 100 fastest SPs each distance, but overall the trend is the same as it is for the AvgSP and MedianSP – Speed Figures increase as distance increases from 4.5 furlongs to 8.5 furlongs (dirt) or 9 furlongs (turf), then begins to decline before leveling out at 16 furlongs.  The main item of note here is that the 100SP at 8.25 furlongs on both dirt and turf is significantly below where the trendline would expect it to be.  The reason for this is likely as noted above: 8.25 furlong races are not carded by the CPU and there were NO stakes run at 8.25 furlongs in 2017.  The lack of stakes races at 8.25 furlongs likely impacts the 100SP data even more than it impacts the MaxSP data; a hotly contested allowance race can produce a very high Speed Figure that gives a high MaxSP (which would technically be an outlier), but you would need 100 of those to match the Speed Figures coming out of the stakes level in the 100SP data.  As a result, the actual 100SP for dirt and turf are approximately 3.2 points and 4 points lower, respectively, than the applicable trendlines.  In fact, removing the 8.25 furlong data would bring the R² value of the trendline for dirt and turf up from 0.9079 and 0.8938, respectively, to 0.9406 and 0.9457, respectively.

Fig. 7 – Dirt and Turf 100SP data separated into sprint races (left) and route races (right).  Red line indicates the average SP of the full data set comprised of 100 fastest SPs at each distance (points above the red line are distances that have a 100SP higher than the overall average; points below the red line are distances that have a 100SP lower than the overall average.

Breaking down the 100SP data a little further (and removing the 8.25 furlong outlier), we can see similar trends in the sprint and route races to the AvgSP and MedianSP data, though the decline in Speed Figures in routes is reduced compared to AvgSP and MedianSP.  For both dirt and turf sprints, Speed Figures increase by approximately 2 points for every additional half furlong in distance, similar to the AvgSP and MedianSP trends.  For dirt and turf routes however, Speed Figures decrease by less than 1 point for every additional half furlong, versus a drop of 1.5 points per half furlong for the AvgSP and MedianSP data.  This data does make some sense.  As noted earlier, not all Sim horses have the stamina for route races, so the AvgSP and MedianSP data is dragged down by the large number of horses putting up subpar Speed Figures at route distances.  As distances increase, fewer horses have the stamina to get the distance, so Speed Figures drop at a high rate (1.5 points per half furlong).  When looking only at the top 100 Speed Figures at each distance, however, we are removing all of the horses that can’t run the full distance and are left only with the horses that are good at those distances.  Therefore, the drop-off as distances increase is less (0.85 points per half furlong for dirt, 0.65 points per half furlong for turf) because the 100SP at each distance is not being dragged down by slow horses.

70SP:

Fig. 8 – Dirt 70SP and Turf 70SP data.  Bubble size indicates the proportional number of races at each distance that had a SP over 70 (the larger the bubble, the more races).  Red line indicates the average SP across all distances after removing all SPs below 70  (points above the red line are distances that have an 70SP higher than the overall average; points below the red line are distances that have an 70SP lower than the overall average).

When eliminating all races with an SP under 70, we see…very similar trends to all of the previous data.  Once again, the trend is for Speed Figures to increase from 4.5 furlongs to 9 or 9.5 furlongs, and then decrease beyond that, leveling out around the 16 furlong races.  Some of the analysis that we saw before is still visible here: 8.25 furlongs is an outlier on the low end (removing the 8.25 furlong data would increase the R² value on both charts slightly); 12 furlongs on the turf is an outlier on the high side, and the R² values for each are pretty strong, at 0.8528 for dirt and 0.9091 for turf.  There isn’t much to add here.

But one interesting note is related to the bubble size.  It’s a little difficult to tell in this chart because the bubbles are small anyway, but the number of races at the extremes (5.5 furlongs or shorter, 13 furlongs or longer) makes up a smaller percentage of the 70SP pool than it did the AvgSP pool.  What this means is that Speed Figures below 70 seem to be disproportionately distributed to the very short (<5.5 furlong) or very long (>13 furlong) races.  For example, the number of 4.5 furlong races on the dirt made up a total of 4.72% of all 2017 dirt races, but the number of 4.5 furlong races of over 70 SP only made up 3.99% of all 2017 dirt races over 70 SP.  This tells us that by cutting off Speed Figures at 70 SP and above, we are losing more races from short sprints and long routes (on a percentage basis) than the middle distance races.  This indicates that it is more likely to run a sub-70 SP in a short sprint or long route than in a middle distance race.  We can see that even clearer in the next chart.

Fig. 9 – Percentage of dirt (left) and turf (right) races that broke the 70 SP barrier at each distance.

These two charts show how many races at each distance cracked the 70 SP barrier.  And unsurprisingly, the trend is similar to our other analyses – the likelihood of breaking 70 SP increases in sprints (from 4.5 furlongs to 7.5 furlongs) and decreases in routes (from 8 furlongs to 16 furlongs), leveling off at the extreme long distances.  At the long distances, this is not very surprising, as (like has been noted previously), many of the horses running those distances do not have the stamina to finish in a fast time.  It is, however, a little surprising to see just how low the percentages at these long distances are: just 50.05% of 16 furlong dirt races were over 70 SP in 2017 (compare this to 8 furlongs, where 90.32% of dirt races were over 70 SP).  It is also surprising to see how low the percentages are in the short sprints.  As noted in the methodology, 2 year olds were removed from the data set; so why would 4.5 furlong races see fewer than 70% of runners break the 70 SP barrier?  Stamina is not an issue at that distance.  There also seems to be a pretty sizeable jump from 6.5 furlongs to 7 furlongs (at 6.5 furlongs on both surfaces, approximately 81-82% of runners break the 70 SP barrier; at 7 furlongs, that number jumps to 87-88%)  One possibility is that these distances do not allow horses to get up to their full speed.  This could be particularly true if the horses running in these races prefer middle distances; they might be slower but have more stamina, which would result in a slower SP at 4.5 furlongs than at 8 furlongs (a horse putting up split times of :23, :23, :23 and :23 would have a slower SP at 4.5 furlongs, but a faster SP at 8 furlongs, than a horse with split times of :21, :22: :25, :28).  Regardless of the reasoning, there is a clear takeaway from this: if you have a horse that can’t crack a 70 SP but want to it to pick up a paycheck, you have a much better chance of doing so in short sprints or long routes than in middle distance races.

Conclusion

So based on the above analysis and results, let’s see if we can answer our initial questions:

1) Do Speed Figures increase with distance?

This is a factual question, and based on the facts, the answer seems to be “yes…and no”.  Looking at five different methods of reviewing Speed Figures for the full data set of 2017 races (AvgSP, MedianSP, MaxSP, 100SP, and 70SP), Speed Figures DO appear to increase as distances increase in sprint races.  But in general, Speed Figures appear to decrease as distances increase in route races.

But one thing to keep in mind here is the old statistical axiom “correlation does not imply causation.”  What this means is that just because the data shows an increase in Speed Figures up to 8f, and then a decrease after that (in other words, that the data is correlated), does NOT mean that the increase/decrease in distance is the CAUSE of the change in Speed Figures.  It’s certainly possible that distance is the cause.  But it’s also possible that there are fewer top-level races at very short and very long distances, and therefore higher Speed Figures gravitate towards the middle distances.  It’s also possible that races at very short and very long distances have fewer runners, which could have multiple effects.  One is that people looking for some easy BPs or placings sneak bad horses into these races just to pick up a 3^rd or 4^th place finish.  Another is that one good horse will dial down its speed during the race because it only needs to be as fast as the next fastest in the race (in other words, a horse than can run a 110 SP at 16 furlongs doesn’t need to do that if the next fastest horse can only put up a 65 SP).

2) Do races at 7.5 furlongs and 10.5 furlongs have inflated Speed Figures?

This is another factual question, and based on the facts, the answer to this one is very likely “no”.  For 7.5 furlongs, looking at each of the five different methods (AvgSP, MedianSP, MaxSP, 100SP, and 70SP), we clearly see that Speed Figures are either in line with, or in most cases below, the expected Speed Figure for such distance (only in Dirt AvgSP and Dirt 70SP is the 7.5 furlong SP in line with the trendline; every other analysis, including all turf calculations, show that 7.5 furlong Speed Figures are below what would be expected).  For 10.5 furlongs, the only calculation method where the Speed Figure is in line with or exceeds the expected SP is in MaxSP (for both dirt and turf).  In every other method, the Speed Figure for 10.5 furlong races is below, and often well below, the trendline showing the expected SP at the distance.

3) Is a Speed Figure at one particular distance “equivalent” to a Speed Figure at a different distance?

This is the most difficult question to answer, and it’s one that our data cannot definitively determine.  What we do know from the data is that if you have a horse that runs a 105 SP at 4.5 furlongs, it will be one of the top Speed Figures at that distance, whereas if the horse runs a 105 SP at 8 furlongs, it will not.  That fact indicates that Speed Figures across distances are actually not “equivalent”.  Instead, the data shown here indicates that a 105 SP at 4.5 furlongs would be equivalent to a 107 SP at 5 furlongs, a 111 SP at 6 furlongs, and a 115 SP at 7 furlongs.  But that isn’t really looking at whether the Speed Figures themselves are equivalent; it is looking at “what is the average Speed Figure at each distance”.  So if you say “My horse ran an average SP”, that would be a 74 SP at 4.5 furlongs on the turf and an 86 SP at 7.5 furlongs on the turf.  Does that mean that the 74 SP at 4.5 furlongs is “equivalent” to the 86 SP at 7.5 furlongs?  Not necessarily – it could indicate is that a horse running an 86 SP at 7.5 furlongs can’t get up to full speed as quickly in a 4.5 furlong race, so the horse’s Speed Figure comes in lower.

So we don’t know from the data if Speed Figures are necessarily equivalent.  We don’t know whether your horse earning a 100 SP at 16 furlongs is “better” than the same horse running a 100 SP at 8 furlongs.  In theory, a 100 SP should be a 100 SP, regardless of the distance.  So whether it’s at 16 furlongs, 8 furlongs, or 4.5 furlongs, a 100 SP should mean the same thing.  But what we DO know from the data is that, whether or not the horse’s ability was the same, the horse is more likely to win with that Speed Figure at 16 furlongs, or at 4.5 furlongs, than it is at 8 furlongs.  We know this because of the analysis above – Speed Figures run in the sim tend to increase in sprints as distances increase, and then decrease in routes as distances increase.  So if you have a horse than can run a 105 SP in 8 furlong races and a 100 SP in 16 furlong races, even though the 8f furlong race appears to be the better/faster race, your horse is more likely to win races at the 16 furlong distance.  That fact alone may be enough to make the question worthwhile.

4) Can I write a BTB article analyzing the data that anyone will read?

Well, if you made it this far, then I guess the answer is yes.

—

Is there another analysis you’re curious about?  Do you see holes in this analysis?  Is there something that you’ve always wondered about the Sim but didn’t have the data to check it?  Let me know if the comments (or on the message board) and I’ll see what I can do.