What is a speed figure/variant?

[Cratos]

I find it interesting that the concept of speed figures are invariably spoken of on the Internet horseracing forums and in some books or magazines, but the statistical or scientific explanations are rarely if ever given.
Shouldn’t the speed figure/variant gurus of the horseracing handicapping world should be able to define the concept?

Also shouldn’t users’ be able to define the tool(s) that they are using?

They don’t have to reveal their methodology because that is proprietary, but an explanation of the applied statistics and science behind their methodology would be informative.

I am not asking for mathematical or statistical models, but verbiage that gives a common sense understanding of speed figure/variant.

[cj]

A speed figure, to me, is nothing more than a better way to express the final time recorded by a horse in a race. That does not mean the horse with the best speed figure was the best horse. We all know there are many things that can influence the final time:


Track configuration and run up
Track surface itself and moisture content
Pace dynamics of each race
Weather, wind
Jockey tactics
Trouble
Ground loss
Track bias
Weight
etc, etc, etc.
Raw time has proven to be a poor indicator of how well a horse ran. A speed figure is a primitive way to try to make the raw time a better indicator of ability. Some figure makers adjust for a few variables, some more. Some are better than others, but they all attempt to do the same thing.

One thing is certain, any halfway competent figure maker can improve upon final time as a predictor of results. This may, or may not, mean the figures are financially better at the betting windows.

[markgoldie]

Beyer gives a pretty thorough explanation of how he makes his figs in his seminal work "Picking Winners" written in 1975.

Essentially what you do is look at the final times for individual classes over a reasonably long period of time. You take the average of these times. This is known as establishing a class par time. The only real trick here is to make sure that some sea change has not occurred relative to the speed of the racing surface. For example, a new track crew might change the depth or composition of surface. And of course, an all-weather installation will drastically affect the par time made for a dirt track. So you want to keep the pars relative to the same basic track surface. Throw out the old ones if there has been a change.

Once you have the pars, you are ready to start making figs. You look at the results of a given card and you compare the final times to the pars. Some will be faster and some will be slower. You find the average difference and that was the raw speed difference for the day. Then you translate that into a final fig number for each class, using a formula for converting fifths of a second into a fig-number adjustment. Not sure what that conversion formula is but it's different for routes and sprints.

Then you may realize that in some races, the result you have come up with makes no rational sense at all. This is considered an abberational number which is adjusted for by a method called "projection." Projection takes a known-quantity horse (like an aged gelding who runs consistent numbers) and projects what his final fig should have been in the race. Using this number as a base, you then assign numbers to the horses who beat him and those he beat based on the number of lengths between them.

Some fig makers may like to project more than just the abnormal races as a way to protect against anomalies.

Some fig makers incorporate trip information in the form of lost lengths into the horse's final fig. Beyer never used to do that and I doubt his numbers are made that way now.

Anyway, in a nutshell, that's what's done in fig making.

Mark

[fmolf]

Beyer gives a pretty thorough explanation of how he makes his figs in his seminal work "Picking Winners" written in 1975.

Essentially what you do is look at the final times for individual classes over a reasonably long period of time. You take the average of these times. This is known as establishing a class par time. The only real trick here is to make sure that some sea change has not occurred relative to the speed of the racing surface. For example, a new track crew might change the depth or composition of surface. And of course, an all-weather installation will drastically affect the par time made for a dirt track. So you want to keep the pars relative to the same basic track surface. Throw out the old ones if there has been a change.

Once you have the pars, you are ready to start making figs. You look at the results of a given card and you compare the final times to the pars. Some will be faster and some will be slower. You find the average difference and that was the raw speed difference for the day. Then you translate that into a final fig number for each class, using a formula for converting fifths of a second into a fig-number adjustment. Not sure what that conversion formula is but it's different for routes and sprints.

Then you may realize that in some races, the result you have come up with makes no rational sense at all. This is considered an abberational number which is adjusted for by a method called "projection." Projection takes a known-quantity horse (like an aged gelding who runs consistent numbers) and projects what his final fig should have been in the race. Using this number as a base, you then assign numbers to the horses who beat him and those he beat based on the number of lengths between them.

Some fig makers may like to project more than just the abnormal races as a way to protect against anomalies.

Some fig makers incorporate trip information in the form of lost lengths into the horse's final fig. Beyer never used to do that and I doubt his numbers are made that way now.

Anyway, in a nutshell, that's what's done in fig making.

Mark
basically a variant ...whether par averaged or projected is a number that purports to tell us how fast a track was in relationship to races previously run at the same class and distance.....most commercial speed figures have the variant already calculated in.If you would like to know the variant beyer is using you need his speed charts and beaten lengths charts and do the math.

[Cratos]

A speed figure is the making of the final times of a horse race conform to a standard by class and in this case it is the par or average time of a given class.

The variant is the amount of deviation from the standard and that metric is use to adjust the final time (called the “raw time”) to the standard. This variant is calculated on a daily basis to reflect the running race times for that day

For you statistics buffs this is called normalization and many fig makers as they are now euphuistically called, have taken this simple process to heights beyond where Andy Beyer was when he popularized this practice in his handicapping ground breaking book, “Picking Winners: A Horseplayer's Guide” which was first published in the mid-Seventies.

However while many fig makers apply a formulaic approach to their speed figures to determine its value for future handicapping I find that disengenious for a variety of reasons.

My primary reason is non-homogeneity, that is a $XX claiming race at Santa Anita is not the same $XX claiming race at Pimilco even though the $XX claiming price is the same. This applies also to allowance races and non-graded stakes races. While graded stakes races (especially Grade 1 races) are the closest to being homogenous they have their class stratifications and this makes any speed figure suspect.

Another of my reasons of doubt about the validity of speed figures is environmental. The weather conditions can influence the times of races very much. There can be great questions about the final time depending upon the wind direction. For instance Belmont Park and Woodbine Turf course offered are the greatest wind impacts because of their track configurations which have large turning radii.

These radii when the wind is speed is over 13 mph have an impact that is felt great on the horse because its body acts like a “sail” when it is negotiating the turns.

There other reasons that I find which make figures suspect, but are too numerous to numerate and define in this post.

However I suggest defining a predictive speed curve for each race and compare it against the historical running curve for that particular class. Why is this better? It is better because two important elements of handicapping would be inherently incorporated into the shape of the curve and they are pace and stamina. Furthermore a confidence interval can be established which would give credence to the horse’s performance.

[cj]

A speed figure is the making of the final times of a horse race conform to a standard by class and in this case it is the par or average time of a given class.

The variant is the amount of deviation from the standard and that metric is use to adjust the final time (called the “raw time”) to the standard. This variant is calculated on a daily basis to reflect the running race times for that day

For you statistics buffs this is called normalization and many fig makers as they are now euphuistically called, have taken this simple process to heights beyond where Andy Beyer was when he popularized this practice in his handicapping ground breaking book, “Picking Winners: A Horseplayer's Guide” which was first published in the mid-Seventies.

However while many fig makers apply a formulaic approach to their speed figures to determine its value for future handicapping I find that disengenious for a variety of reasons.

My primary reason is non-homogeneity, that is a $XX claiming race at Santa Anita is not the same $XX claiming race at Pimilco even though the $XX claiming price is the same. This applies also to allowance races and non-graded stakes races. While graded stakes races (especially Grade 1 races) are the closest to being homogenous they have their class stratifications and this makes any speed figure suspect.

Another of my reasons of doubt about the validity of speed figures is environmental. The weather conditions can influence the times of races very much. There can be great questions about the final time depending upon the wind direction. For instance Belmont Park and Woodbine Turf course offered are the greatest wind impacts because of their track configurations which have large turning radii.

These radii when the wind is speed is over 13 mph have an impact that is felt great on the horse because its body acts like a “sail” when it is negotiating the turns.

There other reasons that I find which make figures suspect, but are too numerous to numerate and define in this post.

However I suggest defining a predictive speed curve for each race and compare it against the historical running curve for that particular class. Why is this better? It is better because two important elements of handicapping would be inherently incorporated into the shape of the curve and they are pace and stamina. Furthermore a confidence interval can be established which would give credence to the horse’s performance.

I certainly don't make speed figures to try conform to any predesignated class level. I also don't try to make class levels at different tracks equal based on price. I doubt many other figure makers do either.

I think you are making a lot of assumptions that are simply not true today, though they may have been in the past.

[Tom]

Who says wind is not already covered? It is.

[Cratos]

Who says wind is not already covered? It is.

How and what is its metric?

[Cratos]

I certainly don't make speed figures to try conform to any predesignated class level. I also don't try to make class levels at different tracks equal based on price. I doubt many other figure makers do either.

I think you are making a lot of assumptions that are simply not true today, though they may have been in the past.

A speed figure does not have to be class related, but it has to be related to some common denominator to make it relevant and this applies to any moving object whether it is horses, cars, airplanes, etc.

There isn’t any way that the class of horses can be set to equality under the present classifying system. Also class is always post determined by performance, but is seldom prior performance implemented probably because of scarcity.

What I am saying is that a horse class will typically be recognized by its performance (sometimes some horses are overlooked), but even if its class is accurately determined there probably would not be enough horses on a given racetrack to always run horses by their determined class and this in part is what accounts for long shots.

To paraphrase Andy Beyer: “there is class within class.”

[fmolf]

A speed figure does not have to be class related, but it has to be related to some common denominator to make it relevant and this applies to any moving object whether it is horses, cars, airplanes, etc.

There isn’t any way that the class of horses can be set to equality under the present classifying system. Also class is always post determined by performance, but is seldom prior performance implemented probably because of scarcity.

What I am saying is that a horse class will typically be recognized by its performance (sometimes some horses are overlooked), but even if its class is accurately determined there probably would not be enough horses on a given racetrack to always run horses by their determined class and this in part is what accounts for long shots.

To paraphrase Andy Beyer: “there is class within class.”
that is the age old question isn't it?...x-class horses will run figures between y and z.....If they run slower they must drop in class to be compettitive if they run faster they may move up in class effectively!i think speed and its measurement are directly related to class

[Tom]

How and what is its metric?

It is part of the "variance."
If final times of 1.12 call for a pace of 46 and I get 47, it is slow 5. If I see a 45, it is fast 5. It might be wind, it might be moisture, it might be too much salt on the popcorn. What it is matters not - how much it is is what counts.

[Cratos]

It is part of the "variance."
If final times of 1.12 call for a pace of 46 and I get 47, it is slow 5. If I see a 45, it is fast 5. It might be wind, it might be moisture, it might be too much salt on the popcorn. What it is matters not - how much it is is what counts.

Wrong and wrong again. Do you watch track and field? The reason I asked is that it has some of characteristics of horseracing and in track and field records are monitored relative to the wind velocity.

Using your example of 1:12 and I assume it is for a 6F race it would a different 1:12 if the race was run at Saratoga and there was a 15 mph NE headwind because that is the orientation of the backstretch and out of the gate the horses would be going against the “resisting force” of the wind and might be more fatigue than normal coming home.

However if it was a 15 mph tailwind from the SW the horses would have the best of it out of the gate and probably have early fast fractions because of the “assisting force “ force of wind.

But if you want to believe that wind force is incorporated inherently, well good for you.

[Greyfox]

Wrong and wrong again....But if you want to believe that wind force is incorporated inherently, well good for you.

I believe Tom is right. Wind is reflected among other factors in the track variant.

If you don't believe we are right, what do you believe contributes to the track variant?

[fmolf]

I believe Tom is right. Wind is reflected among other factors in the track variant.

If you don't believe we are right, what do you believe contributes to the track variant?
i believe you gentleman are right...in a variant usually we are not aware of what causes the variant.it is a simple concept really todays races of a certain class are compared to like races of the same class and distance and the difference is calculated.this difference is the variant and todays speed figures are adjusted accordingly....simple as that!

[Tom]

Wrong and wrong again. Do you watch track and field?



No, I bet horses. And the way I described is more than adequate to predict winners. Nothing else matters but the bottom line. I suspect if you had a better way you would be using it or selling, not arguing it. If you want to think I am losing, then good for you. :lol:

When I get up in the morning and see snow on the ground, I put on a jacket. I have no idea why it snowed, no ide why it got cold, no idea how cold it really is. But just knowing it is some cold a jacket serves my purpose.

[Cratos]

I believe Tom is right. Wind is reflected among other factors in the track variant.

If you don't believe we are right, what do you believe contributes to the track variant?

As I have said earlier in this thread, the so-called “variant” is a deviation from the par or average and therein lays the problem.

The par is calculated over a period of time using final times which are probably not associated with the three wind speeds (tailwind, headwind, and crosswind) that a horse would see during the running of a race.

Given that scenario you cannot say the wind speed is in the “variant” because you haven’t isolated the wind speed in the par to determine its influence and this by no means is an easy calculation.

But to answer your question: “what do you believe contributes to the track variant?” The contributors to the change in the surface speed of the track are many and the primary two are (1) Maintenance and (2) environment.

However the speed of a racehorse will only give a qualitative value to the speed of the track’s racing surface; it will indicate that the surface is slow, fast, or somewhere in between. The reason for this is that you have two variables: the horse internal “motor” so to speak varying due to its performance output and the track surface varying due to what I indicated above. You can not solve a problem for a single answer with all variables.

[proximity]

As I have said earlier in this thread, the so-called “variant” is a deviation from the par or average and therein lays the problem.
.

the start to making a good variant is actually a median of several such deviations from pars or expected performances. and wind speed and direction, while nowhere near even close to being as influential on the race times as it is in track and field, certainly is a factor in horseracing. hell, i'd love to know len ragozin's wind formulas for the different tracks. but even with such information, perfect figures will never be made. that, however, does not mean that "imperfect" figures cannot be useful in a player's handicapping.

[Greyfox]

.

The par is calculated over a period of time using final times which are above. .

If you are calculating variants using pars I would agree.
If you are calculating variants using the best time in the last 3 years I would disagree. It appears to me you are opting for the par method which may be more accurate. The DRF opts for the latter.

[Cratos]

the start to making a good variant is actually a median of several such deviations from pars or expected performances. and wind speed and direction, while nowhere near even close to being as influential on the race times as it is in track and field, certainly is a factor in horseracing. hell, i'd love to know len ragozin's wind formulas for the different tracks. but even with such information, perfect figures will never be made. that, however, does not mean that "imperfect" figures cannot be useful in a player's handicapping.

Because the data that is being collected, wouldn’t the mode be the average that you would want to know?

Incidentally, we started analyzing wind impact in 1978 and continue on to today. What we found out that around the turns at winds above 13 mph the effect can be major, but down the stretch and backstretch not as much because of the “knifing effect.”

[proximity]

Because the data that is being collected, wouldn’t the mode be the average that you would want to know?


i never knew mode to be an average? i always thought it was simply the value that occured the most frequently in a sample of data? so if race deviations from pars or expected performances for a card are fast 2, slow 4, slow 1, slow 3, even, slow 2, fast 2..... then the variant should be fast 2 because that value occured the most frequently on the day? makes no sense to me.

[Cratos]

i never knew mode to be an average? i always thought it was simply the value that occured the most frequently in a sample of data? so if race deviations from pars or expected performances for a card are fast 2, slow 4, slow 1, slow 3, even, slow 2, fast 2..... then the variant should be fast 2 because that value occured the most frequently on the day? makes no sense to me.

As I understand it, the mean is the total divided by n, median is the center value of the string sorted into increasing order and if the string is odd. It is the average of the two middle values if the string is even, and the mode is the most frequent value which works well for the center of the data set when it is non-homogenous

[Tom]

Who says I use the mean?
Or mode?
Or median?

This is NOT a statistics class.
This is horse racing.

Can better variant be made? Certainly.

Is anyone making them?
Maybe, maybe not, but if they are, are they worth it?

If anyone has a better way, post it.
Otherwise, telling people who are doing something and making money off it
is pretty foolish, IMHO.

As far as wind, is anyone measuring it EXACTLY on the track at EXACTLY the time the horses area running at that EXACT spot? If not, your numbers are off.

Meanwhile, I have several nice pace plays coming up Sunday, thank you CJ, even if you have no idea how fast the wind is blowing! :D

[Cratos]

Who says I use the mean?
Or mode?
Or median?

This is NOT a statistics class.
This is horse racing.

Can better variant be made? Certainly.

Is anyone making them?
Maybe, maybe not, but if they are, are they worth it?

If anyone has a better way, post it.
Otherwise, telling people who are doing something and making money off it
is pretty foolish, IMHO.

As far as wind, is anyone measuring it EXACTLY on the track at EXACTLY the time the horses area running at that EXACT spot? If not, your numbers are off.

Meanwhile, I have several nice pace plays coming up Sunday, thank you CJ, even if you have no idea how fast the wind is blowing! :D

I now understand why you are the “Poster Emeritus Pro tempore” on this forum and maybe one day I will be able to make money from wagering on horseracing, but until that time I will keep reading you posts and learning.

[Tom]

Sarcasm is not becoming.
Instead of constantly telling people it can't be done, why don't you offer some solutions?

[Cratos]

Sarcasm is not becoming.
Instead of constantly telling people it can't be done, why don't you offer some solutions?

I am not a sarcastic poster and I never said it could not be done, but I disagreed with the methodology. However for your information I developed the Surface Speed Variant (SSV) and the Projected Pace Variant (PPV) nearly 25 years ago. Also I don’t brag about my winnings nor do I cry about my losing.

[Tom]

Well congratulations.
Sure sounded sarcastic to me.

[Cangamble]

Wind of course is part of the track variant when determining speed figures, but I think most automated and even manual systems do not have knowledge of what the wind is for the day or half a day. The wind will be reflective in the final times.
Because the track variant is based not on one race but many, wind and track surface will automatically be taken care of by the time the variant is established.
In most cases wind evens out anyway. Going 6 furlong, the wind is slightly more of a factor down the backstretch. 7 furlongs on a mile track, wind means a bit more going down the backstretch. If the wind is at the horses face down the stretch in a 7 furlong race, the times for 7 furlongs that day would be faster relative to 6 furlong races on that same day. But that will be noticeable to the person or system coming up with the daily variant.

[Tom]

And what if it dies down during a race of two?

[Cangamble]

And what if it dies down during a race of two?
The same as if the track dries up or gets wetter quickly. Split variants for the day are needed.

[cj]

I am not a sarcastic poster and I never said it could not be done, but I disagreed with the methodology. However for your information I developed the Surface Speed Variant (SSV) and the Projected Pace Variant (PPV) nearly 25 years ago. Also I don’t brag about my winnings nor do I cry about my losing.

By all means, fill us in. No company secrets or anything, but saying you have made them doesn't do much if we have no idea what you are talking about.

[Cratos]

By all means, fill us in. No company secrets or anything, but saying you have made them doesn't do much if we have no idea what you are talking about.

Much of what I have posted or inferred is part of the SSV and PPV methodology with the exception of their proprietary components.

However we don’t sell anything and I am not authorized to advertise on this forum and even if I were I wouldn’t disclosed the SSV and PPV methodologies.

But I will say it is all about inferential statistics and curve fitting for which equations was perfected from the Equibase database going back to 1990. Also we developed algorithms to normalized the circumferences of the racetracks by using the Belmont racetrack as the standard. However in our findings, we found that Pimilco is the near perfect dirt racetrack.

[John]

However in our findings, we found that Pimilco is the near perfect dirt racetrack.

Perfect in what way ? for the horses, for the handicapper and why do horses ship out of Pimilco and win at other tracks. Please explain to me.

[Cratos]

Perfect in what way ? for the horses, for the handicapper and why do horses ship out of Pimilco and win at other tracks. Please explain to me.

When compared to the other racetracks in North America Pimilco turn radii and stretch length was ideal for a horse running either in sprints or routes. The next best racetrack was Aqueduct before the changes to it. Keep in mind I am speaking about the racetrack design not about the horse and bettors.

[John]

When compared to the other racetracks in North America Pimilco turn radii and stretch length was ideal for a horse running either in sprints or routes. The next best racetrack was Aqueduct before the changes to it. Keep in mind I am speaking about the racetrack design not about the horse and bettors.

Thanks, very interesting.

So you are say that although most tracks are one mile in circumference.They are all design differently, Some have sharper curves and others have a longer or shorter stretch.

[delayjf]

Question,

Do churchill and Pimlico have longer than normal stretch runs (for 1 mile tracks) because

a) their turns are tighter
OR
b) because of the placement of the finish line?

I think the answer is b.

[Cratos]

Question,

Do churchill and Pimlico have longer than normal stretch runs (for 1 mile tracks) because

a) their turns are tighter
OR
b) because of the placement of the finish line?

I think the answer is b.

A typical 1 mile track stretch length is 990 feet or less.

Churchill Downs Stretch = 1,234.5 Feet
Churchill Downs Turn Radii Length = 1,292 Feet

Pimilco Stretch = 1,152 Feet
Pimilco Turn Radii Length = 1,326 Feet

[CBedo]

Cratos, thanks for posting your thoughts. With regards to your "perfect" track. Besides for turn radii and stretch length, wouldn't you also need to take into account the bank of the turns, and maybe even the overall "speed" of the track, or am I not understanding your definition of perfect?

Thanks

[socantra]

When compared to the other racetracks in North America Pimilco turn radii and stretch length was ideal for a horse running either in sprints or routes. The next best racetrack was Aqueduct before the changes to it. Keep in mind I am speaking about the racetrack design not about the horse and bettors.
I will agree that Pimlico is a fine racetrack, but even if I accept your designation of it as the "perfect track", just exactly what do I gain from this?

I am not a designer of racetracks. I am a bettor of horse races and most of them are not carded over "perfect tracks".

Of what use is this information to the handicapper/bettor?

[Cratos]

I will agree that Pimlico is a fine racetrack, but even if I accept your designation of it as the "perfect track", just exactly what do I gain from this?

I am not a designer of racetracks. I am a bettor of horse races and most of them are not carded over "perfect tracks".

Of what use is this information to the handicapper/bettor?

Pimilco has turn radii on both turns of almost exactly ¼ mile. Also it has a slight arching, upward curve, or convexity for horses entering the turn because all turns have an effect on the horse because racing injuries are more common near turns, Also with an incline surface before the turn and having a more gradual turn and/or transitional turns reduce injury risks at Pimilco

But you question is about you as a bettor and the effect that its racetrack design has on handicapping. The track has almost four equal ¼ mile segments which allow for good comparison between turn time and straight time. It is a roomy track being 70 feet wide which allows for a liberal 5 foot running path for each horse up to 14 starters.

However with the demise of Magna I am not sure about the current maintenance of Pimilco.