Incomplete Programs

[Jaguar]

Do other folks feel the same way I do that quite a number of very well-thought-out handicapping programs are shipped in rudimentary form?

For example, JCapper- undeniably a brilliant program( I don't own it, but watched a friend use it at OTB). Yet, the program is sold without having been fully optimized. The developer wants you to train it.

Who has the time for this? Most developers have huge databases, from which many track models may be derived.

A software package, particularly one that is fairly costly, should be set up and ready to run right out of the box.

As well, all horse programs should incorporate-at least- robust linear regression- and preferably full-blown A.I.

Some developers seem to be stuck in a time warp. Their desk calendars are
apparently dated "1996".

Jaguar

[Steveh]

I think you've got a great idea but not sure how such a program could be created that would use AI to grind out a +ROI and learn when trends change and adjust. Poly is a good example of a change that the program would have to digest. I believe there was at least one AI program but it didn't seem to survive. I'm not a programer but if you throw some suggestions on how to make such a program, maybe someone will take the challenge.

We have all seen different ideas that have done well for awhile and then the profits fade away. I feel that anything based on what is profitable today will not last forever.

It seems that the people that are able to grind out a steady profit also have to work to maintain what they do. We're all lazy by nature but I don't think software will ever completely replace the need to work at making money in this very tough game.

[PaceAdvantage]

Do other folks feel the same way I do that quite a number of very well-thought-out handicapping programs are shipped in rudimentary form?

For example, JCapper- undeniably a brilliant program( I don't own it, but watched a friend use it at OTB). Yet, the program is sold without having been fully optimized. The developer wants you to train it.Hey Jag, I'd love to know who this mysterious "friend" of yours is at this OTB. You've been talking about him for five years now I believe. Tell him to register an account. I think he'd like it here....

[acorn54]

i don't think you understand the concept behind jcapper
i have jcapper and it is fully functional for what it is suppose to do, provide an ability to query a database and make udms based on ones research.
jeff might want to elaborate since he is the developer.

[hcap]

As well, all horse programs should incorporate-at least- robust linear regression- and preferably full-blown A.I.

Jaguar
I don't think A.I. will work-at least not now. My experience is it tends towards averages with large data bases as the basis for the training model, and jumps into imaginary trends using short term models.

I have however found an occasional profitable situation going forward, but IMHO, for every profitable going forward there may be an equal or greater number that are disasters financially. The problem may be knowing which to act on and for how long? Even withholding a "set" for testing against models "found" is lacking. Benter did not use A.I.

[garyoz]

As well, all horse programs should incorporate-at least- robust linear regression- and preferably full-blown A.I.

Some developers seem to be stuck in a time warp. Their desk calendars are
apparently dated "1996".

Jaguar

Linear regression didn't work in 1996 and won't work in 2007.

[betchatoo]

Do other folks feel the same way I do that quite a number of very well-thought-out handicapping programs are shipped in rudimentary form?

For example, JCapper- undeniably a brilliant program( I don't own it, but watched a friend use it at OTB). Yet, the program is sold without having been fully optimized. The developer wants you to train it.

Who has the time for this? Most developers have huge databases, from which many track models may be derived.

A software package, particularly one that is fairly costly, should be set up and ready to run right out of the box.

As well, all horse programs should incorporate-at least- robust linear regression- and preferably full-blown A.I.

Some developers seem to be stuck in a time warp. Their desk calendars are
apparently dated "1996".

Jaguar

If JCapper (or any program) were sold ready to be a "black box," and worked, it's useful shelf life would be a month. Soon prices would be driven down and the horses picked would be underlays. If you don't have time to train and use your program in your own way (which is another term for handicap) then you should pick another game to play.

[jandrus]

I know the person who jags is talking about. I belive he scans this board.

[richrosa]

Linear regression didn't work in 1996 and won't work in 2007.


Almost entirely true.

The trick is to understand the trends of the past, and apply it to the race that is unfolding in front of you. For example, there are certain rankings that lend themselves into being very predictive, especially when they happen in concert with each other and the race tilts towards the fitting the running style of the entry.

Like many others, I spend a lot of time trying to disqualify false favorites in an effort to bridge the gap over the takeout. A linear regression can aid in this respect in the areas of comparison to those dominant predictive factors, or more importantly trends at a track, distance, and surface that will work against a favorite.

So in a nutshell, linear regression is a powerful tool, but it is not likely lead you into the promised land without true interpretation and careful analysis.

[DanG]

Almost entirely true.

The trick is to understand the trends of the past, and apply it to the race that is unfolding in front of you. For example, there are certain rankings that lend themselves into being very predictive, especially when they happen in concert with each other and the race tilts towards the fitting the running style of the entry.

Like many others, I spend a lot of time trying to disqualify false favorites in an effort to bridge the gap over the takeout. A linear regression can aid in this respect in the areas of comparison to those dominant predictive factors, or more importantly trends at a track, distance, and surface that will work against a favorite.

So in a nutshell, linear regression is a powerful tool, but it is not likely lead you into the promised land without true interpretation and careful analysis.
Very well said Rich!!! :ThmbUp: :ThmbUp: :ThmbUp:

[njcurveball]

A software package, particularly one that is fairly costly, should be set up and ready to run right out of the box.




You are very "book smart" in this statement. If everyone is doing duplicate work, why not incorporate that in the parent program and propogate to all the children.

Unfortunately or perhaps fortunately for us, this is a pari-mutuel game and not one where the person who picks the most winners makes the most money.

You are thinking more like this is a vegas game where the odds are set and do not change.

However, if you really like this deterministic approach you could easily get free selections at a number of web sites.

Heck, just keep watching here. There are some guys who give mortal locks from time to time!

[DanG]

If everyone is doing duplicate work, why not incorporate that in the parent program and propogate to all the children.

Unfortunately or perhaps fortunately for us, this is a pari-mutuel game and not one where the person who picks the most winners makes the most money.

You are thinking more like this is a vegas game where the odds are set and do not change.

However, if you really like this deterministic approach you could easily get free selections at a number of web sites.

Curveball bringing the ‘Jaba heat!!! :ThmbUp:

http://upload.wikimedia.org/wikipedia/en/d/d3/Sopr_Bada_Bing1.jpg

[ryesteve]

For example, JCapper- undeniably a brilliant program( I don't own it, but watched a friend use it at OTB). Yet, the program is sold without having been fully optimized. The developer wants you to train it.
You clearly don't understand the program. It doesn't require "training". The concept of "training" isn't even applicable to Jcapper. And there is no "optimization" that's lacking. You can either use the reports it generates to handicap, or use it as a research tool to develop your own spot plays. If you were actually a user of the software, perhaps you'd have a better idea of what it does and how it works.

[Lefty]

JCapper allows you to put in your own factors from drop down menus and create and test your own personal methods. Or as RyeSteve said, you can just use the reports and handicap from the numbers.

[hdcper]

For me, Jcapper is the program of choice!

Jaguar, you may want to request a copy of the demo from Jeff and see for yourself. No cost and an opportunity to give it a test drive.

If you decide to take a look, feel free to PM me and I will be glad to walk you through the basics, to jump start your learning curve.

It is a great program and everyone should take a look.

Bill

[JustRalph]

don't ever start a discussion of Data analysis with the HTR boys......... :lol:

They will eat you alive!!! and Good for them!!

[Jeff P]

JCapper is not artificially intelligent. It doesn't "learn" as more and more races are fed into it. It is what Acorn54, Betchatoo, Ryesteve, and Hdcper all say it is. It will take your Bris or TSN data and results files and build a database at the click of a button.

JCapper comes with a unique set of factors and numbers - most of which are derived from algorithms of my own creation arising out of thousands of hours of hard research into the results of large data samples... research that I performed myself. A few of the factors are my own rough approximations of good ideas read about in books or taken from user suggestions that have proven effective when I tested them out. Included in the factor set are compound ratings representing early and late pace, class, ability from speed figures, the competency of the horse's human connections, current form, several different power ratings, and a handful of different yet accurate probability assessmet algorithms.

The 2007 program also has an interface that gives the user the ability to name and weight factors of the user's own choosing and roll the result into the user's own custom power rating and odds line.

JCapper has a module called the Data Window which allows the user to not just query a database - but to slice and dice it almost any way he or she wants to see horses broken out in terms of the program's unique factors and numbers.

The user has the ability to save queries that show promise roi-wise as UDMs and build on them. IHMO, this is the single most important aspect of the program. In my own case it was the concept of UDMs that enabled me to turn the corner from being a losing player and become a winning player.

Many people have asked me how I handicap. My answer almost always surprises them. 99 pct of my handicapping isn't done on race day. It is done well beforehand - at the Data Window. My work is saved as UDMs. On race day, JCapper does what it was designed to do: It very quickly shows me those horses that qualify as UDM plays so that I can focus on them - and ONLY them. I am a highly disciplined player. I know from experience that my UDM plays are profitable. Viewing actual results of thousands of UDM selections each year over the years has taught me one thing: If I do only one thing and nothing else - if I bet my UDM selections and bet nothing else - I become a winning player.

Creating UDMs that perform well going forward is not easy. But it is a skillset that can be learned. And if it can be learned, then I believe it is something that can also be taught.

The user isn't expected to fly blind in this endeavor. There are help docs and podcasts that explain the entire learning curve... from operating the basic program, to benchmark tests of the program's factor set, to the theory behind creating viable UDMs. A couple of months ago I wrote a help doc called Finding an Edge. In that help doc I tried to take a lifetime of experience and lay out exactly how and why my own methods of creating UDMs can be replicated by the user to create his or her own UDMs that will perform well going forward. The doc contains a step by step example of how to use the Data Window and UDM Wizard to create such a UDM using specific JCapper factors.

No, the program does not come fully optimized. I naively tried that once and was not happy with the outcome. The original program (Fall 2004) came with two simple UDMs (one primary factor/two secondary factors) that were profitable right out of the box (black box style) with no player decision making whatsoever. Once I sold enough copies - and the avg mutuel of the selections of those two UDMs began to suffer - I pulled the original program from the market. This happened in early March 2006.

I spent the next year playing professionally (using JCapper as my only past performance/software tool) while simultaneously improving the program itself.

JCapper2007 contains an expanded factor set and has capabilities far beyond those found in the original program. In the right hands - especially for the user/player willing to spend some serious time at the Data Window and who is willing to WORK at improving his or her game - the program can provide a very nice edge.

Which brings me to the real reason I responded to the original post in this thread...

IMHO, handicapping and betting thoroughbred races to produce steady consistent profits is something that is very hard to do. I would be lying if I said otherwise. I've spent half a lifetime figuring out how to do just that. And I've spent about 17 years morphing the original program that I started out with into what JCapper is now... IMHO it is a VERY good tool - especially for the player willing to put in the time.

No, the program isn't optimized. That's actually by design. You see I wanted to create a powerful tool that can be used by the player in a number of different ways. No matter what handicapping style the player likes to focus on... speed and pace, trainer patterns, track profiles, form cycle, class, breeding, identifying playable races, accurately identifying vulnerable and strong favorites - the program does have the ability to help the player improve in each of these areas... The basic underlying philosophy behind JCapper is that the user player who is willing to out-work, out-research, out-create, out-innovate his or her competition is the user player who deserves the rewards.

To my way of thinking that is the only way a handicapping database program can have any type of shelf life in a parimutuel world.


-jp

.

[CapperLou]

Do other folks feel the same way I do that quite a number of very well-thought-out handicapping programs are shipped in rudimentary form?

For example, JCapper- undeniably a brilliant program( I don't own it, but watched a friend use it at OTB). Yet, the program is sold without having been fully optimized. The developer wants you to train it.

Who has the time for this? Most developers have huge databases, from which many track models may be derived.

A software package, particularly one that is fairly costly, should be set up and ready to run right out of the box.

As well, all horse programs should incorporate-at least- robust linear regression- and preferably full-blown A.I.

Some developers seem to be stuck in a time warp. Their desk calendars are
apparently dated "1996".

Jaguar

Hello Jag:

JCapper sure as heck is not stuck in a 1996 or any time warp--You have written to me at times and indicated you hope to come to FL during a Gulfstream meet--please do--and we'll see what it does for me the WAY I USE IT when you are here.

I have scaled back my play because of poly surfaces, but even though I'm making half the plays I used to and even skipping some days that I normally play--I'm still showing a profit based on patience (lots of it) and waiting for the right plays to appear. The hardest part is patience.

And, unlike many other JCapper users, I do not have many UDM's. This is a remarkable piece of software and you know I've seen most of them.

I would try it if I were you--if not--I would go to HSH or HTR and if you use your knowledge and pick your fair value plays etc with any of these three programs--I think you will be happy with what happens!!!

CapperLou

P.S. Hey Lefty, if you know of any nice foreclosures in LV--please PM me.So many, it's hard to know where to start looking!!!!

[Jaguar]

Admin: Excellent idea. But, I think he wants to remain a lurker. He runs a busy retail business and has a family. In his off hours, he is busy training JCapper.

My guess is he doesn't want to spend time communicating online, which is too bad. Alot of us could benefit from his input.

I used to get to OTB every day, for years. But, now my wife and I are caregivers for my 95-year old Dad, and I can't get out the way I used to. So, I only see my handicapping friend a few times a year nowadays.

I am planning on buying JCapper with my January check, God willing, and will probably want to put my two cents in- concerning the program- around that time.

From what I've seen of JCapper, Jeff must really be an incredibly smart guy.

Jaguar

[Jaguar]

Acorn 54, in my opinion, Jeff has written an absolutely smart, right on the money horse racingprogram. I hope to buy the program in a few months.

Jeff's unique methodology finds profitable kernels of information in the great stew of vast amounts of horse racing data. As Jeff explains, this on-going analysis requires a great deal of time and testing.

I am 64 years old and have been training and testing horse and greyhound computer programs since Tandy made their first tape-storage computers in the early 80's.

My goal in life is to find a horse handicapping program which uses neural nets and genetic programming and which trains easily and quickly, making superb performance and financial(betting) models.

In all candor, I have to say that Jeff's output is so brilliant that I really wouldn't mind putting in the training time. I only saw JCapper used one afternoon at OTB, but it knocked me out.

Jaguar

[Jaguar]

garyoz, I find your comment interesting and have one question for you: if you are correct, why does Multicaps work so well, even without neural nets and genetic programming?

Admittedly, the program at this point in time, is dated and badly needs to be re-done, but it is certainly the most effective linear regression program I have seen.

Jaguar

[Jaguar]

Jeff P, many thanks for your very clear, easy to understand, explanation of how your excellent program works. I learned more from your post than from your website.

As an old man, I am probably stuck in the mud regarding handicapping methodologies, and your very sensible approach to the puzzle is a real wake-up call for me.

Jaguar