Parts of the racing form that you ignore?

[podonne]

Writing some code to produce my own past performances and I started to notice that several parts of the form don't seem to be relevant anymore. Curious which parts of the form that could be improved as far as your handicapping process is concerned.

Some of mine:

• Earnings at various types of races (lifetime, prior year, current year, fast dirt, track) - Thinking purses aren't as reliable any more
• Weights in the "first three finishers" part of the past performances
• Range of claiming prices in a particular prior race, as opposed to the claiming price of the race. Claiming price of the horse in the race might be better?
• Milliseconds in the fractional times
• The race number of a prior race
• Owner? Does the owner really matter?
• The rank of a work on that day, but still probably whether it was a bullet work

Also curious about the repeated Starts - Wins - Places - Shows parts for jockeys, trainers, at different race types. Would it be easier just to see Starts, Win%, ITM%, ROI?


Any other thoughts?

[Overlay]

I realize that purse amounts and structures can be affected by factors that are not strictly related to quality or performance, but I still find average earnings-per-start (as originally calculated by Quirin) to be a useful ranking indicator that encompasses both class and consistency.

[DeltaLover]

Pretty good posting and question!

For my systems, the most important thing I am ignoring from the pps is the name of the horse itself!

Sound a bit strange but let me explain it...

I have gone through a long process of what and how to present in my past performances... Years ago I've started writing pretty facny 'fat' desktop applications mimicking is many ways the classical look of a pp as appears in the racing form adding quite a few improvements... Displaying data in graph formats, all the matches of the horse for my handicapping factors with their associated IV and ROIs, creating every possible combo of factor per trainer per jockey per surface, displaying run against each other in a graphical representation and many many more...

After a pretty long evolution curve of my software, today it looks totally different than it used to be... I try to display as little information as possible...

I am sure you can see where I am getting to... Displaying just the number of the horse and the amount I am going to bet.. I have not gotten there yet but at least I am in the point to present each starter in a race with an array of numbers (close to 10) that are sufficient for me to make a decision. None of this numbers can be found in the pps as appear in the racing form, in contrary the are the output of my models that are receiving the 'primitive' data converting them to indexes giving me a clear representation of each starter in a single line. Any kind of descriptive data in a textual format, like horse, trainer or track names for example are not part of this view which is a matrix containing as many rows as starters and as many columns as the indexes I am using. This format of data is very easy to back test using betting simulators and even if we need to form an opinion using our judgment we can be sure that any bias influencing our decisions will be easy to quantify and eliminate.

So:

NO HORSE NAME
NO TRACK CODE
NO TRAINER NAME
NO JOCKEY
NO SURFACE
NO DISTANCE
NO RACE CLASSIFICATION

and the list goes on....

[JohnGalt1]

I use the Bris pp generater so print the pp's the way I feel most useful.

I don't print Bris's speed number or their pace numbers. I print only four calls, 1c 2c str fin. I also don't print the company lines. And I print only the last five workouts.

I do print the Daily Racing Form SR and Track Variant, but I make my own pace figures. And I print the purse of each race.

I use 28 days as my lay off line not the pre-set 45 days.

I do better with simplicity, and reduce clutter.

I know some programs can look at 20, 30 or even more factors. If I did use that many, I'd either be so confused in a race and not bet, or if most of the factors favored one horse, I'd be betting chalk.

[HUSKER55]

Deltalover: Nice post! FWIW you just pinted out some errors I am making and I thank you.

Horse's names for example. You are right. Data either fits the matrix or it does not.

[DeltaLover]

Originally Posted by HUSKER55

Deltalover: Nice post! FWIW you just pinted out some errors I am making and I thank you.

Horse's names for example. You are right. Data either fits the matrix or it does not.

A common pitfall most of the serious students of the game are falling in has to do with confusion in the layers of decision making process. We should start from the bottom which consists with the very primitive data adding an analytical tier which will transform them to comparable numerical quantities and then continue with synthetic processes until all the noise associated with individualities of each starter is eliminated have as lean and comprehensive representation of the event.

Using simulations we can easily conclude of the validity or our views, so if we need more detailed data we add them to the base 'primitive' level and start again until we have a viable approach. Starting with an increased quantity of detailed data should be avoided since the more details we have the larger our sample needs to be and the higher is the probability to reach wrong conclusions.

[Robert Goren]

I tend ignore the lengths back at any call other than finish and first quarter. even then I take the first quarter lengths back with a grain of salt. I just trust the numbers at the other calls.

[HUSKER55]

DeltaLover, can you give me a simple example of what you mean by "primitive data" and "analytical tier"?

[DeltaLover]

Sure...

I call primitive data all the raw data :
for example:
fractional times, final time, finish position

The should be passed through an analytical level that will transform them to normalized quantities that will serve as the basis for the next processing tier(s). For this level we can either use custom trasnformation or third party numbers.

example:
Track variant, speed figures, pace figures , quirin points etc.

In top of this layer we build higher level synthetical layers converting the second tier numbers to statistical measurements (like for example the average SF of the best performance of each starter and its STDEV might present a valuable metric

Then in top of these we build our betting model..

I think you can understand this process better if you read this posting in my blog:

http://alogatas.wordpress.com/2012/1...f_early_speed/

[bob60566]

Raw data

For me the most underlooked is the dates the horse contested. How many people break it down to days and not just look at the date the horse ran.

[therussmeister]

Whilst you aren't interested in the rank of the workout, I'm not interested in the time of the workout. I use rank to judge workouts. I learned this at Sportsman's Park one year when the track was dreadfully slow with 5 furlong works generally going in 1:05 and change. Rank was the only way for me to judge accurately.

Also, since I make my own speed and pace figures, I could theoretically dispense with all times whatsoever, but I wouldn't. It just wouldn't seem right.

[therussmeister]

Originally Posted by bob60566

Raw data

For me the most underlooked is the dates the horse contested. How many people break it down to days and not just look at the date the horse ran.

I break it down to days, I count exact days up to 2 months, then I just count months.

[DeltaLover]

Originally Posted by bob60566

Raw data

For me the most underlooked is the dates the horse contested. How many people break it down to days and not just look at the date the horse ran.

Besides the fact that thedrf track variant is systematically based by day of the week, I do not know how the date a horse ran can add any original value.. With this I mean that a race in Belmont stakes day will already have a high class rating based in the class of the contenters and adding the date as a factor will only add redundant info ..

Unles you mean something different.. Can you please clarify?

[Capper Al]

I'm due to revamp my system. The output is displayed in Excel with about a dozen or more tabs. Tabs for speed, class, form, etc. A lot of it is repetitive. For example, name and Morning line odds appearing on most sheets. About the only things that is not being used is owner.

[bob60566]

Originally Posted by DeltaLover

Besides the fact that thedrf track variant is systematically based by day of the week, I do not know how the date a horse ran can add any original value.. With this I mean that a race in Belmont stakes day will already have a high class rating based in the class of the contenters and adding the date as a factor will only add redundant info ..

Unles you mean something different.. Can you please clarify?

Yes
With the lower class horses the form cycle is very easily seen and comeback races within ceetain days between races beforse a good race and when layoff is comong up that to me adds to the original value when you see days instead of just months. With class horses this is not relevant but at lesser tracks when horses have no workouts for months at time and race into shape ypu soon learn the form cycle.