One idea is to divide the career prizemoney by the number of career starts, which gives the API, then multiply this by the place strike rate. This means that if the horse won three good races in a row as a 2 year old, but has very average form recently, this equation drags down the API to a more reasonable level. Not perfect, but it's an idea.

__________________
RaceCensus - powerful system testing software.
Now with over 413,000 Metropolitan, Provincial and Country races!
http://www.propun.com.au/horse_raci...ng_systems.html
*RaceCensus now updated to 31/01/2025
Video overview of RaceCensus here:
http://www.youtube.com/watch?v=W821YP_b0Pg

It's what Elle MacPherson has and Kylie Minogue doesn't.It is defineable and at the same time a darned elusive quality to capture.It is that certain "something" that is there, but not tangible.

In racing it tends to manifest itself after the event.Class horses perform in quality events.A horse that has previously run in a class3 handicap and is running today in a class1 event does not make it "the" class runner.

Forget class,it will drive you crazy !!

Hi all,

AL ILLICH, once said:

"Horses of superior class defeat horses of inferior class because they have superior speed. Yes, speed and nothing but speed. Not necessarily speed as expressed in the final time of a race, but speed at the right time, speed when it is most needed. ...In order to understand why this is so, it is necessary to understand the importance of the pace of the race, meaning the speed prevailing at different stages of a race."

a. CLASS IS MORE IMPORTANT IN LONGER RACES.
b. CLASS IS MORE IMPORTANT IN HIGHER Classes
c. EVEN HIGHER CLASS Horses START OUT IN MAIDEN RACES.

A Good Class Indicator should produce: a. The highest percentage of WINNERS in its HIGHEST RATED Horses, and the lowest percentage of WINNERS in its LOWEST RATED Horses; b. A higher percentage of WINNERS in HIGHER Classes, and its lowest percentage of WINNERS in the lower Classes; c. A higher average percentage of WINNERS in longer races than in shorter races; and d. A very high percentage of WINNERS in MAIDEN GRADE races.

In short, to me this can only be accomplished by analysisng past winners sectionals and finding where the "right" speed of the race is applied.

Good one Woof43, but keeping it as vague as possible doesn't clarify the problem all that much. We know that the fastest horse "on the day" has won the race after the fact: obviously, that horse had the speed at "the right time" since it WON. What we'd like to know is how you know IN ADVANCE that that nag that's always coming home "like a rocket" in every race is NOT going to do so today or, conversely that the 100-1 shot WILL (and be on it!!!!). We would like predict that "maybe" (as in the 2000 Olympic 1500m final) the world record holder,who hasn't lost in 4 years, is going to get nutted by a guy who's best time EVER is 5 seconds slower. What use can it possibly be (from a punting point of view) to sit there, only able to mutter "he had the speed AT THE RIGHT TIME"!!!! What next with this info,Woof??

Hi Punter 57,
as i previously said,
"A Good Class Indicator should produce: a. The highest percentage of WINNERS in its HIGHEST RATED Horses, and the lowest percentage of WINNERS in its LOWEST RATED Horses;"

Using prizemoney doesn't achieve the above.

By using the appropriate Speed/Sectionals it WILL provide a smooth transition thru the rankings.
As I also previously said by analysing past winners sectionals you will find which sections or segments of the race are more predictive of the winner then others.


"We know that the fastest horse "on the day" has won the race after the fact: obviously, that horse had the speed at "the right time" since it WON. What we'd like to know is how you know IN ADVANCE that that nag that's always coming home "like a rocket" in every race is NOT going to do so today or, conversely that the 100-1 shot WILL (and be on it!!!!)."


The use of the Standard Deviation function comes in very handy in the above comments an in your 1500m Olympic Final example, it is also one of the most powerful functions to use in racing, if its used correctly you can develop a speed performance envelope for each runner based on its past performances

The performance envelope will be a Quick Time an a Slow Time, but before you do that you need to find the Standard Error in your sample size(each individual horses LTD Past Performances) which is STDev/Sqrt(Number of starts).
Once you have that figure you can then apply it firstly to find the Slow Time
you would use the following formula; Avg LTD Time+(1.96*STd Error figure) the reasoning for using 1.96 most performers running 1.96 STDev's slower then their avg LTD would lose.

Now to find a Horses Quick Time you basically do the same as above but you need to substitute the 1.96 with another figure, if you look at past quick times in a Horses previous form you will find a much lower figure then the 1.96 STDev's when a Horse has run its best performances.

The simplest thing to do once your achieved the above is to combine each Horses StDev with the above Times an run them thru a Monte Carlo simulator which will provide you with a good set of probabilities...

Morning Woof43. If you have time (I'll keep an eye on this thread) could you, perhaps give an actual example of how you would do that for tomorrow's race (any tomorrow!!). This is NOT a challenge or anything like that as I'm aware that nothing is INFALLIBLE. Whether it pans out the way it "should" on any individual race is not so important as the long term principle. I'd much appreciate it as you really seem to have a fine grasp of the maths/stats. Thanks

woof43,
as Vic. Provincials and most country tracks Aust. wide don't record sectionals what do you do there?

Woof43 - In your example, how would you take into account the daily track variant? By that I mean that a track could be rated Good but could be running fast on the day (better side of good), or slow on the day (slower side of good and close to Dead). That's of course if the track condition is correct in the first place (and we know that a lot aren't). There are numerous occasions when a number of either track, race or class records are broken on the same day/same meeting, which is a clear indication of the daily track variant that is fast on the day and therefore the times have to be treated with caution.

Hi La Mer

The simplest way to calculate variants is to start by creating a set of average winning times for every distance(segments)/Class combination at a track. Then after each meeting is finished, take the actual winning time for every race on the card, and divide it by the average winning time for that distance(segments)/Class. Then I take the average of those numbers. That gives you the variant for the day. It should be a number very close to 1.000. If the number is below 1.000, it indicates the track was slower than average that day. If greater than 1.000, the track was fast. Then, you take the finish time(sectional) for every Horse that ran that day, and you multiply that finish time(sectional) by the daily variant.
Once this is applied to each runners LTD performances his average will have a lower standard deviation than an average of regular times without a Track Variant, if you've done everything correctly.

To have good results with the variant-adjusted times, you must be a bit careful with your distance(sectional)/Class averages, and your database maintenance. It's important to create a good set of standard averages. If you ever change your standard average winning times, you must re-process your entire database and recalculate the daily variants and re-apply the new variants to every single past performance line in your database. Many people mess things up by changing their standard averages and not reprocessing all of their historic data.
Testing your procedure for creating variant-adjusted times is very easy. Pick out about 20 Horses at random from your database, each of which has at least 20 past performance lines. Calculate the standard deviation of each Horses raw times, and of their adjusted times. If a strong majority of your 20 Horses (say 15 of them) have a lower standard deviation on the adjusted times than the raw times, then you are probably doing things okay. If 10 or less of the Horses have a smaller standard deviation on the variant-adjusted times, you probably have a problem.
This is one procedure where it's not possible to "over-fit" the back data. The better you minimize the standard deviation, the better your handicapping results will be.
If your thinking of using any times other then winners times, i have completed numerous studies on placgetters etc The standard deviation for Class X winners was .151951 seconds. The standard deviation for Class X second-place finishers was .157107 seconds, which is about .005 seconds greater than for the winners. So this analysis confirms for me that putting place finishers into a variant adjustment will increase the resulting runners standard deviations versus using only winners and thats not our goal.

The other issue of track degradation does have some effect but it can work in both directions, The above variant as it been described is quite good at removing daily variations in the racing surface. But if you want it to be even better, you could apply a secondary variant. Take the average of the averages, and then adjust each race for the difference between the average for that race number (over a long time period) and the average of the variants. That would give you a daily variant that was adjusted for track degradation as well.

Woof43 - Thanks very much for your explanation. although it looks like a lot of hard yards, but probably well worth the effort. I wish you well in your endeavours in this area.