Why has technology left handicapping in the dust?

[headhawg]

Originally Posted by DanBoals

Have you tried ChatGPT? Ask it who will win an upcoming race. Or, try GPT4All or SuperAGI or some other software in that vain you can run on your local machine, see how it does.

ChatGPT and like language processing tools are not designed to pick winners of horse races. People need to read up about machine learning and genetic algorithms before making comments about "AI" and horse racing.

[PaceAdvantage]

Originally Posted by DanBoals

What is the incentive for this?


If you are Fidelity or TD Ameritrade or someone else like that, you get money when your customer trades by screwing them on the price differential. If you charge them 10 cents more a share than "market" you make a fortune on the millions of trades per day. Also, you look at your customers that are making money over time and build a model based off their trades and then you let them do the brain-work and you profit mimicking their trades.


As stated elsewhere, if you have handicapping software that really works, why on earth would you sell it? Look at the NHC, 4 days work for $800k grand prize, all profit. If you were a developer and your software was really great, you would win this competition every year. I do not recall a software vendor ever winning this or any other major competition.


Have you tried ChatGPT? Ask it who will win an upcoming race. Or, try GPT4All or SuperAGI or some other software in that vain you can run on your local machine, see how it does.

You sound like you're accusing me of asking for a black box.

I'm sorry if I conveyed that impression, but that's not what this thread is about.

[Dave Schwartz]

Originally Posted by DanBoals

Truer words are seldom spoken.


I think most of us love figuring out who the winner will be and put little to no effort in figuring out how to make money on that knowledge.

Are you Dan Boals the college fullback?

[DanBoals]

Originally Posted by headhawg

ChatGPT and like language processing tools are not designed to pick winners of horse races. People need to read up about machine learning and genetic algorithms before making comments about "AI" and horse racing.

I have not used ChatGPT because of the lack of privacy you have to sign away to ask it questions. SuperAGI and GPT4ALL can be downloaded. I am quite sure you have no clue what GitHub is, but their source code is at the following links:

https://github.com/nomic-ai/gpt4all

https://github.com/TransformerOptimus/SuperAGI


They are not a black box. It would require a bit of work to get them set up and analyzing horse races. For someone who knows nothing about software engineering, this would probably be a daunting task.


None the less, you probably will never, ever be able to create your own neural network, so this is the closest you are going to come. Machine learning is a pretty generic term that can cover this field, but a genetic algorithm is probably not what you think it is. The idea that you are going to turn loose a genetic algorithm on something vague like find me a race winner and have it develop magically to do so is pretty ridiculous if you actually understand what a genetic algorithm is.


Ideally, you would write your own neural network being a top 1% software developer and then being an expert handicapper, you could fill in pretty reasonable values for the initial variables and then let it run over your dataset for a few months or more and it would point you in new directions.


Since you do not even know the difference between an AI and an expert system my good Mr. Hawg, I would suggest you take my advice and use these two programs as a starting point. These tools have already been used to design course curriculums, pick stocks better than all but a couple funds and write scientific papers. That fact that you have no idea what this software does in no way limits the software.

[DanBoals]

Originally Posted by PaceAdvantage

You sound like you're accusing me of asking for a black box.

I'm sorry if I conveyed that impression, but that's not what this thread is about.

Not sure what that means...


SuperAGI and GPT4ALL are generic expert systems. This means if you put the work into learning them, and then you feed them with good data and then you set them to work finding what you are interested in learning from that data, they will do that for you.


Writing a neural network that does this in a more specific way would be a very time consuming task for even the best software developers in the world, although I believe it would be a more accurate solution to delivering maximum return on investment. You might never get to the point of expertise to do that. That guy over in Hong Kong took this path with his software guys and seems to have been successful from the article I read on him where he talked about having something like 250 values to consider in the decision.



These software systems are able to do in a generic form what you want, or at least what I think you were saying you wanted, which is given a dataset, to find things you tell them to look for.


This is kind of the opposite of what I think of as a black box, maybe we just have different definitions of what black box means.

[DanBoals]

Originally Posted by Dave Schwartz

Are you Dan Boals the college fullback?

No. I am Dan Boals the veteran and retired software developer who is living in Reno and trying to enjoy his retirement.


I have no athletic ability, but having been born and raised in Oklahoma, I am versed in the religion of football and played from 5th thru 7th grade at which time I think I weighed about 90 lbs and most of the guys I was lining up against were between 160 and 240. In 7th grade, I think we had one offensive lineman that was over 6' 5", our own Shaq of the farm league.


Ahhh.... to be tall and athletic, that would have been nice.

[headhawg]

Don't be a condescending asshat Mr Dan. Of course I know what github is. I have already been to the links that you posted. I never claimed to be an expert but using ChatGPT as an indicator of what AI is in the context of handicapping is amateurish at best and clueless at worst. And you have no idea what I am capable of doing.

[DanBoals]

Originally Posted by headhawg

Don't be a condescending asshat Mr Dan. Of course I know what github is. I have already been to the links that you posted. I never claimed to be an expert but using ChatGPT as an indicator of what AI is in the context of handicapping is amateurish at best and clueless at worst. And you have no idea what I am capable of doing.

PaceAdvantage, if I understand correctly, funds this board out of his own pocket and maintains it on his own time. That is a very nice thing to do for the handicappers who visit here. Not a lot of people do very nice things for others. You were giving him very bad advice. I felt that needed to be corrected.


I could have uploaded him several books on how generative AI works but that did not seem productive. Unless someone really just loves software, reading book after book on a new software technology is pretty boring. Having him ask who will win, then waiting for the race to run, and then asking why the software made the choices it made, seemed like a much better solution to getting a non technical person to understand what steps the software went thru to come to its conclusion and to see if those steps were logical to him.


With ChatGPT, you get access to an enormous amount of hardware, so it can "learn" things super quick compared to the two software programs I recommended for him. If he absolutely hates working with that kind of technology, he can find very quickly with very little time invested thru ChatGPT. But, if he sees value in how that program works, he could then decide he wanted to install and train his own software with his own data and then use that to test his ideas and maybe learn new ones with that software on his own machine, no one else in the world will have the same analytical tool he has since he is the one that fed it the data and he is the one that told it what to do.


Having a unique piece of software that is looking thru your own tracks for your own angles seemed to be what he wanted. I could be way off on that and if I am I apologize. This would save him from having to hire professional developers and pay a huge overhead, which when he referenced equity and derivative analysis systems is what I thought he wanted.


When you told him that generative AI could not do this, it seemed pretty obvious, you did not understand the technology. I had a friend in 1996 write a neural network for tuning the PWM loops in our amplifiers which was my intro to neural nets. I started using them in 1997 to model and adjust factory automation systems, so I have been working with them for over 25 years now and for me, that would be my approach. But for someone not technical, that is a huge learning curve with no guarantee of success, hence, my recommendation to use ChatGPT to see if you like the idea and then if you do, to download and create your own analysis system using open source versions of generative AI.


Calling that very logical approach amateur and clueless demonstrates which person is the clueless amateur.

[PaceAdvantage]

headhawg, you need to chill a bit. This is not a thread I want turned into a name calling festival.

Dan, thanks for your responses. I think there was some misunderstanding at first, but I see where you're coming from now.

[Nitro]

Originally Posted by PaceAdvantage

headhawg, you need to chill a bit. This is not a thread I want turned into a name calling festival.

Dan, thanks for your responses. .

I certainly second that motion!
Thank you, Dan, for taking the time to provide us (non-software engineers) with your very nice and very well-articulated explanation of the Machine programs as related to AI. To be honest I’m not a big fan of AI, not only because of its obvious limitations in this environment, but because of the things you mentioned about the type of resources required to even implement such a program.

I believe AI has proven itself in many of the financial markets. To me “financial” means Money. Now I can’t say for sure, but the person who developed the tote analysis program I use may have in fact been able to establish and build in many of the idiosyncrasies and betting psychologies into to this program.

I know these types of comments are counter intuitive to many of those who follow standard handicapping techniques who think those methods can be turned into some sort of AI program. They seem to ignore the realistic changing physicality of the participants and the motivations of those controlling every aspect of their racing careers,

Enjoy your Retirement!
.
.

[Dave Schwartz]

Originally Posted by DanBoals

No. I am Dan Boals the veteran and retired software developer who is living in Reno and trying to enjoy his retirement.


I have no athletic ability, but having been born and raised in Oklahoma, I am versed in the religion of football and played from 5th thru 7th grade at which time I think I weighed about 90 lbs and most of the guys I was lining up against were between 160 and 240. In 7th grade, I think we had one offensive lineman that was over 6' 5", our own Shaq of the farm league.


Ahhh.... to be tall and athletic, that would have been nice.

You live in Reno and we've never met?

Why don't we change that?

[TonyK@HSH]

Originally Posted by DanBoals

Writing a neural network that does this in a more specific way would be a very time consuming task for even the best software developers in the world, although I believe it would be a more accurate solution to delivering maximum return on investment. You might never get to the point of expertise to do that. That guy over in Hong Kong took this path with his software guys and seems to have been successful from the article I read on him where he talked about having something like 250 values to consider in the decision.

Dave Schwartz, also living in Reno, created a Neural Network around 1990. I felt it was his strongest software.

Tony

[classhandicapper]

Originally Posted by PaceAdvantage

You sound like you're accusing me of asking for a black box.

I'm sorry if I conveyed that impression, but that's not what this thread is about.

I may have said this earlier, but I think what you really want is a database with all the PP and result information available so you can do research that answers all the handicapping questions you have and possibly find some potentially profitable areas.

That's sort of what Formulator has for trainer data, but you want to extend it to everything.

That's why I built my database starting in 2014.

Learning by trial and error or pen and paper is way too time consuming to get the large samples you need to verify something.

I have to tell you though. I've learned an amazing amount of stuff from my database. I feel like I am no long operating with theories and ideas on a lot of things. I have definitive answers. But I've found very little that's sustainably profitable in an automated way. At this point, I know way more, can make more accurate odds lines, and construct tickets better, but it still comes down to carefully analyzing the race and waiting for races where I disagree with the public and feel confident I'm the one that's right long term. Even then, it's not easy because there are constantly things coming that are so particular to a specific horse or race, you can't really know what the long term expectation is.

[DanBoals]

Originally Posted by Dave Schwartz

You live in Reno and we've never met?

Why don't we change that?

Hi Dave,


I use Ted's RDSS software, which I really like and usually bet thru the Atlantis on their mobile app, but I plan on changing that. The Atlantis RaceApp has been giving me problems for the last month or so, so I think I am going to start going in and betting at the casino.


I played Saratoga last year and it and Keeneland were my most profitable tracks. If you would like, I would be glad to meet up with you at any casino as long as they have an outlet I can plug my laptop into like I can at the Atlantis. I used to be able to handicap off the DRF, but these days I need my software.


My email is simply my username at Yahoo if you want to get in touch. Last year, it seemed like the first two weeks were the easiest to profit off of, so I plan on starting betting on Thursday the 13th of July. If you are going to be in town then, shoot me an email and we will meet up. I have a couple comp dollars at the Atlantis so if we meet there, I would gladly use them to buy you lunch or dinner.


Dan

[Dave Schwartz]

Originally Posted by DanBoals

Hi Dave,


I use Ted's RDSS software, which I really like and usually bet thru the Atlantis on their mobile app, but I plan on changing that. The Atlantis RaceApp has been giving me problems for the last month or so, so I think I am going to start going in and betting at the casino.


I played Saratoga last year and it and Keeneland were my most profitable tracks. If you would like, I would be glad to meet up with you at any casino as long as they have an outlet I can plug my laptop into like I can at the Atlantis. I used to be able to handicap off the DRF, but these days I need my software.


My email is simply my username at Yahoo if you want to get in touch. Last year, it seemed like the first two weeks were the easiest to profit off of, so I plan on starting betting on Thursday the 13th of July. If you are going to be in town then, shoot me an email and we will meet up. I have a couple comp dollars at the Atlantis so if we meet there, I would gladly use them to buy you lunch or dinner.


Dan

I, too, am playing at the Atlantis.
You are right - the app is horrible.
Making more than WPS or a simple exacta ticket is nearly impossible.

I was forewarned that there would be times when I'd get shut out for no apparent reason. Haven't had that yet, but admit I've only tested the interface.

While there are other casinos who use apps funding the accounts needs to be in person. Sadly, I couldn't find any other casinos in Reno with an app.

The good news about the Atlantis is that, if you bet there on a steady basis, you will earn lots of comp points. The bad news is that one can only eat so many seafood buffets. LOL


I'll reach out via email.
Look forward to discussing AI, among other things.