PDA

View Full Version : Race Factor Pairs (or what to do when you're snowed in)


Ray2000
01-06-2018, 07:07 PM
The Chart shows the effect of selecting a horse with 'Best in Race' with 2 paired factors.
Example, if you bet a horse which has the fastest speed rating and the best driver in the race . .
then you can expect a strike success of 36.6% and a ROI of .878.

Best Win% is no surprise, it's the horse with best Trackmaster speed and best on tote board (49.2%).
Best ROI goes to Trackmaster Speed and Best starting position. (0.954)

(Ignore Best Final time coupled with Best 4thQ time . . .they are way 2 dependent on each other.)

The surprise to me is the Driver/Trainer combo.
I expected these horses would be over bet so a ROI of 0.713 is OK
but the strike rate is strange, 24.0%


Data covers 19777 races for 2017 but the sample size for finding a horse with any 2 bests, is around 4000

coachv30
01-06-2018, 10:02 PM
Great Data Ray!!:ThmbUp:

Bogo
03-14-2018, 01:33 AM
The Chart shows the effect of selecting a horse with 'Best in Race' with 2 paired factors.
Example, if you bet a horse which has the fastest speed rating and the best driver in the race . .
then you can expect a strike success of 36.6% and a ROI of .878.

Best Win% is no surprise, it's the horse with best Trackmaster speed and best on tote board (49.2%).
Best ROI goes to Trackmaster Speed and Best starting position. (0.954)

(Ignore Best Final time coupled with Best 4thQ time . . .they are way 2 dependent on each other.)

The surprise to me is the Driver/Trainer combo.
I expected these horses would be over bet so a ROI of 0.713 is OK
but the strike rate is strange, 24.0%


Data covers 19777 races for 2017 but the sample size for finding a horse with any 2 bests, is around 4000


Would it help or hurt the ROI if we made it best TM speed in last two or three?

Ray2000
03-14-2018, 08:08 AM
The speed numbers used in this test were the average of the horses pplines. Break lines, qualifiers, or lines older than 30 days were not used. This usually gives an average of 3 lines. I never tested the highest TM speed number of the last 3 but most of the time it would raise the speed rating for all horses in the race, and not change the horse with 'Best in Race' factor selection.

Dave Schwartz
03-14-2018, 11:20 AM
The speed numbers used in this test were the average of the horses pplines. Break lines, qualifiers, or lines older than 30 days were not used. This usually gives an average of 3 lines. I never tested the highest TM speed number of the last 3 but most of the time it would raise the speed rating for all horses in the race, and not change the horse with 'Best in Race' factor selection.

Ray,

This is a great way to look at data.

What I've found in data modeling is to build what is called a "Reynolds Number."

(He was a 19th century English physicist who specialized in fluid mechanics.)

I wrote about Reynolds and his numbers in this 20-year old article: Here (https://pacemakestherace.com/who-was-osborne-reynolds/)

Even though the article is old, we've continued to move the technology forward. Today it is literally the core of our handicapping success.


Dave Schwartz

Ray2000
03-14-2018, 02:22 PM
Dave

That's an interesting approach to combining factors.
Have you looked at weighting the Reynolds ranks with impact values?
Perhaps Early speed is more or less important than class.


The Win % impact values for the individual factors posted above were.
(13% Win base used for IV)

4.7 Morning Line Pick
3.0 Ave TM Speed
2.3 Ave My speed
2.2 Earn$
2.1 Class
2.0 Driver
1.9 Classdrop
1.9 Best Final Time
1.9 2Yr W%
1.7 Trainer
1.6 Best Post
1.6 Early Speed
1.4 Best 4th Q Time

Dave Schwartz
03-14-2018, 03:44 PM
Dave

That's an interesting approach to combining factors.
Have you looked at weighting the Reynolds ranks with impact values?
Perhaps Early speed is more or less important than class.


Absolutely. :ThmbUp:

There are two sides to the coin:
1) Using an absolute number as in the original article.
2) Weighting the best factors together into (what we call) an "Object."

#1 - Contender Selection
When the horse is in the winners circle, and you ask yourself, "What indicated that this horse might be the winner?" - The answer we get most often is that the horse will have one of the following Reynolds number combinations:

11, 12, 13, 22

Notice that 14 (which would be a Reynolds number of four) is not in the list.

Of course, the more factors you consider, the easier it is for a horse to qualify. We have found that the optimum number of factors to use is five.

This process will produce a huge number of winners among your contenders. Furthermore, the great majority of winners that are left out of this process are not gettable with conventional handicapping techniques. In other words, you are not going to get too many $85 horses this way.


#2 - Factor Weighting
As a general rule, unless you have a very large database, any waiting process that you create will be heavily flawed. Even with our large database (around 450,000 races at all times) the proper weight for a given factor will change from one type of race to another.

We face the same issues as everyone else, even with BIG DATA.

Our solution to the problem has been to use a simple, Fibonacci weighting system.

100
62
38
24
15

In this way, all we have to do is determine the ORDER of importance as opposed to the actual proper WEIGHT.


Dave

Bogo
03-14-2018, 06:57 PM
Wow. This is some interesting stuff. Thanks to Ray and Dave. If we were to use 5 factors in a harness race , what are everyone's thoughts on what they would be?

Dave Schwartz
03-14-2018, 08:08 PM
Wow. This is some interesting stuff. Thanks to Ray and Dave. If we were to use 5 factors in a harness race , what are everyone's thoughts on what they would be?

Bogo, I know you are not talking to me because I have not got a single clue about harness racing.

Glad you got something from my response.

Dave

Ray2000
03-14-2018, 08:53 PM
The 5 factors may not be the same for different size tracks or even between tracks of the same size.
And Dave's right when it comes to sample size, breaking it down by tracks usually results in large error margins.



I'll post some numbers for ROI and Win% tomorrow. Tonight's a Hockey night, GO PENS!

Tom
03-15-2018, 11:10 AM
Great thread, guys - really interesting way of looking at things.:ThmbUp: