beginners questions about pace figures

[dilanesp]

I have some questions about pace figures that maybe the more mathematically inclined folks here might be able to help me with.

Up until the mid-1990's (roughly when I got out of law school) I made my own speed and pace figures. I used Beyer's method for speed figures, and made what was basically a Beyer figure for the first half mile of 5 1/2 furlong, 6 furlong, and 6 1/2 furlong sprints. I found the pace figures very useful in identifying when a speed horse was faster than his rivals. I can remember some significant tickets I cashed doing that.

When I no longer had time to make my own figures, I started using the pace numbers out of Today's Racing Digest, again only in the sprints. Again, I've had some big scores using that, the most memorable probably being Trinniburg in the BC Sprint.

Here's my basic, beginners' questions. There's now been an explosion of pace figures. There are Equibase pace figures in the programs that they produce, calculated precisely taking into account the various run-ups. The Racing Form now runs a timeform pace number (just one for each horse, not every race). And Formulator features Moss Pace Numbers. I know Timeform sells pace numbers directly too.

Does anyone have any thoughts on the following questions:

1. My use of pace numbers, as you can see from my discussion, is pretty limited. I just use them to see if there's a superior speed horse in a sprint. What are some handicapping angles beyond this. How should I approach pace numbers in routes, or elongated sprints? Do they have any usefulness in turf sprints? Turf routes? When is a superior pace number irrelevant?

2. Which pace numbers, of the commercially available ones, are the most useful for these purposes?

Thanks in advance for all discussion.

[PressThePace]

I'm not an authority by any means but I look for a couple things that have served me well over time. First, when a horse is up on an exceptionally hot pace while losing ground into the first turn(especially in route races) and still runs credibly, but may fade late...that is a horse that will get my attention. I will always refer to the charts to make sure others weren't able to do the same thing in the race, because there are times when the figures may not tell the whole story. Conversely, if a horse closes into a slow pace to get up or just miss, again merits consideration next time out. Of course, other factors (namely, class for me) come into play. I don't consider anything a given in this game.

[DeltaLover]

Originally Posted by dilanesp

Does anyone have any thoughts on the following questions:

1. My use of pace numbers, as you can see from my discussion, is pretty limited. I just use them to see if there's a superior speed horse in a sprint. What are some handicapping angles beyond this. How should I approach pace numbers in routes, or elongated sprints? Do they have any usefulness in turf sprints? Turf routes? When is a superior pace number irrelevant?

2. Which pace numbers, of the commercially available ones, are the most useful for these purposes?

(1). The best way to discover how to use a pace figures (or any other numerical metric) is to develop an automation mechanism to test many randomly created decision trees and select the most appropriate for the specific race to bet (obviously, the implementation of this system is not trivial and most likely you do not have the required knowledge and talent to accomplish something like this).

(2). The usefulness of any figure is a product of its effectiveness, so your question is incomplete as you need to define the “cost function” or in other words how you are comparing different methodologies among themselves. For example, a less accurate figure can very well represent a superior metric compared to one that fits the data curve closer, assuming that we can discover a deviation threshold, beyond of what we can catch huge betting anomalies.

[dilanesp]

Originally Posted by DeltaLover

(1). The best way to discover how to use a pace figures (or any other numerical metric) is to develop an automation mechanism to test many randomly created decision trees and select the most appropriate for the specific race to bet (obviously, the implementation of this system is not trivial and most likely you do not have the required knowledge and talent to accomplish something like this).

(2). The usefulness of any figure is a product of its effectiveness, so your question is incomplete as you need to define the “cost function” or in other words how you are comparing different methodologies among themselves. For example, a less accurate figure can very well represent a superior metric compared to one that fits the data curve closer, assuming that we can discover a deviation threshold, beyond of what we can catch huge betting anomalies.

Delta- no way can I do (1) , you are quite right. And I do understand why (1) is necessary to really determine the effectiveness of the numbers.

Not sure what (2) is getting at though.

[whodoyoulike]

Originally Posted by dilanesp

I have some questions about pace figures that maybe the more mathematically inclined folks here might be able to help me with.

Up until the mid-1990's (roughly when I got out of law school) I made my own speed and pace figures. I used Beyer's method for speed figures, and made what was basically a Beyer figure for the first half mile of 5 1/2 furlong, 6 furlong, and 6 1/2 furlong sprints. I found the pace figures very useful in identifying when a speed horse was faster than his rivals. I can remember some significant tickets I cashed doing that.

When I no longer had time to make my own figures, ...

1. My use of pace numbers, as you can see from my discussion, is pretty limited. I just use them to see if there's a superior speed horse in a sprint. What are some handicapping angles beyond this. How should I approach pace numbers in routes, or elongated sprints? Do they have any usefulness in turf sprints? Turf routes? When is a superior pace number irrelevant? ...

I don't think it's that limited. I always felt the 4f mark was very important but I do consider the other fractional calls for race scenarios (i.e., shape). I actually don't handicap races less than 5.5f. Races from 5.5f on up reports the 4f call. I think what you're missing as far as what you've posted is applying or referencing it to a database and it's relationship to the race shape in general using stratification of your database because "class" is different for each distance.

[DeltaLover]

Originally Posted by dilanesp

Delta- no way can I do (1) , you are quite right. And I do understand why (1) is necessary to really determine the effectiveness of the numbers.

Not sure what (2) is getting at though.

OK, I will try to explain #2 in a few words (although I am not sure that I am articulating my thoughts very well).

Let’s assume that we have some methodology to estimate the accurateness of a specific metric that measures some attribute like a pace figure. This methodology can be a very analytical one that considers very low level and primitive data, like the geometry of the track, the humidity of the track, the wind direction etc or it can be any other model that provides us with enough confidence that we can estimate the mechanics of a race with a high accuracy. Please note that I am not claiming that this methodology really exist and I only use it to explain my point.

Now, let’s consider two different algorithms to estimate the pace figure and we want to decide which of the two is better.

What I am trying to say, is that even if one of the two might be closer to the “real” value as dictated by the analytical model, the other one can very well be superior for betting purposes.

if we can detect some patterns by the application of an additional layer of indirection to reform it, the more “wrong” figure might reveal many more positive EV outliers and become the basis for a superior betting approach assuming that is only used when a significant (amplified) value is calculated.

[cj]

I'll try to add more later. For now, I'll say that applying an adjustment based on run up distance is very, very risky for a variety of reasons. If horses were always gunned out of the gate like quarter horses, it would be fine. But thoroughbreds, even in sprints, are not.

[dilanesp]

Originally Posted by DeltaLover

OK, I will try to explain #2 in a few words (although I am not sure that I am articulating my thoughts very well).

Let’s assume that we have some methodology to estimate the accurateness of a specific metric that measures some attribute like a pace figure. This methodology can be a very analytical one that considers very low level and primitive data, like the geometry of the track, the humidity of the track, the wind direction etc or it can be any other model that provides us with enough confidence that we can estimate the mechanics of a race with a high accuracy. Please note that I am not claiming that this methodology really exist and I only use it to explain my point.

Now, let’s consider two different algorithms to estimate the pace figure and we want to decide which of the two is better.

What I am trying to say, is that even if one of the two might be closer to the “real” value as dictated by the analytical model, the other one can very well be superior for betting purposes.

if we can detect some patterns by the application of an additional layer of indirection to reform it, the more “wrong” figure might reveal many more positive EV outliers and become the basis for a superior betting approach assuming that is only used when a significant (amplified) value is calculated.

I'm seeing it now, thanks. One of the many subsets of "look for factors that give you positive EV, not necessarily factors that identify the most likely winner of the race".

[Tom]

I found the pace figures very useful in identifying when a speed horse was faster than his rivals. I can remember some significant tickets I cashed doing that.

With this in mind, start looking very hard the new TFUS early pace numbers.
You will find many, many situation where they will point you to good priced winners.

Case in point....the Oaks last Friday.
Many people were conceding the race to Paradise Woods off her powerful front running win in the last race. The Early pace figs suggested a fast pace and if you believed that, you easily has the winner, the top LATE pace horse at 9-1. And the icing on the cake, the #3 LP horse ran second at 12-1.

This has been going on for years. Cj introduced those figs to his clients a long time ago and they have stood the test time.

Remember, they are not race figes, they are horse figs - they use several recent races not just one. Pick your spots where you see certain horses may be favored by the pace what are nor favored by the toteboard.

They are there, they are easy to use, try them, take some notes.
Discuss live races here......best research is done live with real money. Those lesson tend to stick more.

[Tom]

At NYRA, run up = grab up.

[thaskalos]

The set of pace figures is incomplete if it doesn't also provide a rating for the race's final fraction. Only by consulting BOTH ratings can the horse's overall performance be properly evaluated.

Another major function of the accurate pace ratings is that they sometimes signal the impending improvement (or deterioration) of the particular horse's "form", thus allowing the horseplayer to ANTICIPATE the fluctuations of form that we so often see...instead of just expecting the horses to "repeat" some recent race(s) that they've already run. The bigger profits lie in the "projection"...IMO.

For instance:

An unusually-quick half-mile rating is often a "wake-up" factor for a horse, even if it leads to an unimpressive final effort resulting in a slow final speed figure. And an unusually-quick final-fraction rating often is an indication of a poorly-timed ride by the jockey...suggesting that the horse was capable of displaying more early speed in the race, if the jockey were so inclined.

Many people use "pace-handicapping" in order to predict how the upcoming race is likely to be run, but, IMO, the most useful aspect of pace-handicapping is that it allows us to better understand how impressive the PRIOR races of these horses were. Only by thoroughly understanding the PRIOR races can we form an accurate opinion of what the future may bring.

[DeltaLover]

Originally Posted by thaskalos

The set of pace figures is incomplete if it doesn't also provide a rating for the race's final fraction. Only by consulting BOTH ratings can the horse's overall performance be properly evaluated.

Another major function of the accurate pace ratings is that they sometimes signal the impending improvement (or deterioration) of the particular horse's "form", thus allowing the horseplayer to ANTICIPATE the fluctuations of form that we so often see...instead of just expecting the horses to "repeat" some recent race(s) that they've already run. The bigger profits lie in the "projection"...IMO.

For instance:

An unusually-quick half-mile rating is often a "wake-up" factor for a horse, even if it leads to an unimpressive final effort resulting in a slow final speed figure. And an unusually-quick final-fraction rating often is an indication of a poorly-timed ride by the jockey...suggesting that the horse was capable of displaying more early speed in the race, if the jockey were so inclined.

Many people use "pace-handicapping" in order to predict how the upcoming race is likely to be run, but, IMO, the most useful aspect of pace-handicapping is that it allows us to better understand how impressive the PRIOR races of these horses were. Only by thoroughly understanding the PRIOR races can we form an accurate opinion of what the future may bring.

Great posting!

[CincyHorseplayer]

Originally Posted by thaskalos

The set of pace figures is incomplete if it doesn't also provide a rating for the race's final fraction. Only by consulting BOTH ratings can the horse's overall performance be properly evaluated.

Another major function of the accurate pace ratings is that they sometimes signal the impending improvement (or deterioration) of the particular horse's "form", thus allowing the horseplayer to ANTICIPATE the fluctuations of form that we so often see...instead of just expecting the horses to "repeat" some recent race(s) that they've already run. The bigger profits lie in the "projection"...IMO.

For instance:

An unusually-quick half-mile rating is often a "wake-up" factor for a horse, even if it leads to an unimpressive final effort resulting in a slow final speed figure. And an unusually-quick final-fraction rating often is an indication of a poorly-timed ride by the jockey...suggesting that the horse was capable of displaying more early speed in the race, if the jockey were so inclined.

Many people use "pace-handicapping" in order to predict how the upcoming race is likely to be run, but, IMO, the most useful aspect of pace-handicapping is that it allows us to better understand how impressive the PRIOR races of these horses were. Only by thoroughly understanding the PRIOR races can we form an accurate opinion of what the future may bring.

The final fraction figure is a must. Any compound rating will not suffice. Especially when evaluating turf performance. I don't see a true late pace rating anywhere except in the numbers HDW puts out. Figure maker variants are their true virtue. But this rating is more important than even that IMO.

[dilanesp]

I should have mention sustained pace numbers, which I do use on turf races. I learned about them from one of Beyer's books. Is it better to just look at a figure for the last fraction than to use a number that takes into account the entire race?

[DeltaLover]

Originally Posted by dilanesp

I should have mention sustained pace numbers, which I do use on turf races. I learned about them from one of Beyer's books. Is it better to just look at a figure for the last fraction than to use a number that takes into account the entire race?

The real problem is to define what "better" means and how it is tested and confirmed something that is related to my previous posting in this thread.

Ideally you need to be in the position to decide upon the value of any kind of metric without knowing what (or how) it is measuring.

Your first and most important step must be to transform the descriptive data into a vector containing numerical and categorical data in such a way that no significant information is lost but also unnecessary noise is minimized. Take as an example the following horse whose past performances are described using my proprietary metrics and software:



You probably recognize some of the data but you have no idea what each metric means of how it is calculated, the idea is to be able to decide whether some (or all) of them are "good" for betting or prediction purposes.

Even if you do not know what each number means, you still can select some of them and reduce the dimensionality of the data to something like the following matrix:



You follow the same procedure for all the available horses in the race and you end up with an array of similar matrices.

Now, assuming that you have a full historical data base you can do the same for every available race and using its outcome (including the finish sequence ,the odds or even the fractional calls if you like) you are missing a black box that will decide weather your input data (the matrices you created before) contain sufficient data to create a winning selection system or not. If you cannot find such a system you need to improve your existing metrics or create new ones.