DOWNLOAD HERE - Newest Version of HandiFast

[mikesal57]

Originally Posted by headhawg

I don't know if Handifast is "wrong" as it's not designed to be a data analysis program. If you want to use R then use it on the raw data. Proprietary factors like Bris PP and the like already use some weighted combination of data elements. I am by no means a stats wiz, but it seems to me that running a regression on variables that are not completely independent doesn't really give you the key to the mint. This is especially true if you don't know what data goes into those factors. So the analysis might find five great factors but that's because they all contain a common data point, like 2nd last speed rating.

When and if you get into regression analysis you will see your questions are answered working with the program....

Theres 2 parts to it....first is Correlations.....

With correlation , you'll find out which factors have the same relationships...
then you can eliminate them

Then theres regression that will tie up all the factors to winners and give your a weighted value for each factor....

This is what the whales did in their early stages....now they have

[Handiman]

Here is the question. If there is no scaling and a factor is weighted say 100 points. How do you determine which horse gets what for that factor? The other thing is, the difference between the factor raw numbers like Jockey win percent and Best Lifetime speed for instance. Jockey win might be 30 and best lifetime speed is 95. How do those factors compare in importance? How do their numbers compare?

Mike, if you are running down the right road, with all of the research you have done, I'd think you would have a list of the most important factors down to the least. If you make them available, I will for sure put them in Handifast.

If there is no scaling how do the factor points get distributed?

Handi

[acorn54]

hi handiman, hope you are in good spirits. at the risk of being an uninvited guest, i would hope i could shed some light on this conversation. basically, you are asking mike to share his "finished" product, or magic formula if you will.
inevitably, in a general way you are dealing with the well trampled factors, but in a delicate balance, superior to how they are reflected in the mutuel pools, in the combination you are,you are using, but if it is a positive roi for you it is profitable, usually by just a few percentage points, as the R2 is pretty low nowadays for the betting markets.

[mikesal57]

Originally Posted by Handiman

Here is the question. If there is no scaling and a factor is weighted say 100 points. How do you determine which horse gets what for that factor? The other thing is, the difference between the factor raw numbers like Jockey win percent and Best Lifetime speed for instance. Jockey win might be 30 and best lifetime speed is 95. How do those factors compare in importance? How do their numbers compare?

Mike, if you are running down the right road, with all of the research you have done, I'd think you would have a list of the most important factors down to the least. If you make them available, I will for sure put them in Handifast.

If there is no scaling how do the factor points get distributed?

Handi

To answer your first question.....simple...take raw number X weight ...done

Second question.....Maybe like Brian mentioned prior...Index the raw number to range of 100 to 60...

ExAMPLE: Jockey raw numbers are between 30 and 4....make 30 = 100 ....4 becomes 60.......all the other fall between them

Best Lifetime Speed...Top 95.....Bottom 46......Set 95 at 100....set 46 at 60....the rest fall between them...

[mikesal57]

As to important factors within HandiFast......

A friend has learned how to get regression within Python....its called VIF

Results....it pretty much reflects my findings too

[Ghostrider5]

Don't know if this is a factor. But horses that have a combination of wins or lost a race by a length or less. Divided by the number of past performance races showing in racing form. You will get multiple horses for this.

Also, horses that rank in the top four speed figures in two or more of their last four races. Compared to other horses in the race.

And horses that improved in position or beaten lengths or both in last race. Compared to the second race back. The has hidden trainer moves. With all of the above, you will get multiple horses. But good prices since you are not looking at obvious factors.

One more. Take early pace. And add best two late pace or last fraction figures. Maybe take top four horses with fastest feet per second. Two furlough faction with those horses best two late pace figures.

Ghostrider5

[Handiman]

Thank you guys for the input. Mike I am assuming that the larger the number the better.

Ghost, I will certainly look at those. In this context they would in fact be considered factors.

Acorn, Mike and I have had the factor discussion a lot. He mentioned Quality over quantity. I wanted to give him a chance to provide that quality to Handifast if he wanted too. The idea behind version 13 is to give the user more flexibility in factor use. Some will be good and some not so good as is everything in life. And since the basic program has always been free and will continue to be so as long as I am coding, I invite anyone to offer up suggestions.

Whether Mitchell's indexing is better than another way of quantifying raw numbers I'm not sure. Nor am I sure of the integrity of raw number times factor points to arrive at a factor total.

I wish we had a statistics guy to shed light on this. But may be both become an option. Then it would be up to the user to decide. I will see how much work the coding will be.

Handi

[Handiman]

There is a problem that has popped up in my mind. If we were to do as Mike suggest and take raw factor figures and multiply it against the factor points, if using 22 factors the total point numbers could become huge. To big to even display. As I envision 13 as of now, user will have the option of selecting anywhere from one to 22 factors to put in each factor set. So the aforementioned action would be viable in a factor set with 5 or 6 factors. But huge with a set of 22. Mitchell indexing was mentioned. I suppose each factor could be indexed as such and no scaling used. Just add the indexed factor values up for total points. In either way not sure how the fair odds would be affected.

Looking for some input.

[headhawg]

Mathematically speaking, you could divide your total by some constant - like 1,000 -- and it would not affect anything. You would just be moving the decimal point. Or divide each factor by the same number. If you don't want a zillion decimal places then just use a rounding function. In fact, you could probably just truncate to an integer after the division.

[Handiman]

I have thought about that and have left it as a possibility. I am trying to get some conversation from people who view this thread and may indeed have a stake in the outcome as users and potential users. I thank you buddy for giving my thoughts legitimacy concerning the scaling down to a displayable number. It is, if anyone is not aware of, very costly to have someone write a complicated handicapping program for them. As a result, not many do. Here is a chance for handicappers to offer up suggestions as to what they might want in a program. At least for their views to be seen and considered at no cost to them. Oh and speaking of costs. I would like to thank those that have seen fit to donate, as I have been able to pay a couple of medical bills resulting from my 12 day interment in the ICU ward this last year. Still have quite a bit to payoff. But anything surly helps. Thanks again Guys.

Handi

[headhawg]

Originally Posted by Handiman

I have thought about that and have left it as a possibility. I am trying to get some conversation from people who view this thread and may indeed have a stake in the outcome as users and potential users. I thank you buddy for giving my thoughts legitimacy concerning the scaling down to a displayable number. It is, if anyone is not aware of, very costly to have someone write a complicated handicapping program for them. As a result, not many do...

As you know, I am working on my own programs. However, feel free to reach out to me if you need some help. And yes, programmers are expensive. Thanks for your contributions, and also thanks for giving me motivation every time you post about Handifast.

[Handiman]

Thanks Buddy. Still use your code. And may in next version offer user chance to index all factors vis a vie Mitchell. And then just add the numbers together for point totals. We'll see how that affects the fair odds. They'll either be good or bad...

Handi

[Hues10]

Hi there. Anyone have a problem loading up a Brisnet file for Gulfstream or Parx? The file is recognized, but it does not show any horses, and the program shuts down. Very strange.

[mikesal57]

Originally Posted by Hues10

Hi there. Anyone have a problem loading up a Brisnet file for Gulfstream or Parx? The file is recognized, but it does not show any horses, and the program shuts down. Very strange.

You could have gotten a corrupt file...

rarely did anyone have this issue

[Hues10]

The program has worked fine with Saratoga and even DelMar. Just strange that two other tracks did not work. OK. Thanks for the response. Wanted to see if it was the Program possibly.