Ai - Pace and Cap

Old Arkie Gal · 05-20-2024, 05:03 PM

I have been away but I haven't seen anything posted about AI and the potential
effect on handicapping. I don't know much about AI or its applications but it seems it might have a deleterious effect on it.

DanBoals · 05-21-2024, 12:53 PM

Hi Old Arkie Gal,

Are you from Arkansas? I grew up outside of Tulsa, so same area if you are...

The only person I know trying to sell an AI system is Dave and he talks about what he has at: https://www.community.horsestreet.com/discussions/p1

From what I see and having talked to him in person last summer about it, it isn't as good as RDSS from my standpoint.

AI is really a marketing term. What people are calling AI are actually what programmers have called Expert Systems for well over 30 years. The AI's being touted use what we call Neural Networks, which means they vary constants and weighted variables slightly and see what the impact is on results. Neural Networks have been around for about 30 years and I used them in factory automation and motion control software in the late 90's. Howard even tried them but did not have the technical knowledge to do much with them and so rejected the idea. What is called generative AI, like ChatGPT is basically using an expert system trained thru a neural network that has criteria set by the developers and is focused on conversing with users. When they give results like black nazi soldiers in the german army in WW2, it is a hilarious example of how bad they are.

What we call AI will improve and may someday actually be AI. Right now, the problem people have with AI in a field like racing is that the programmers do not understand handicapping enough to tell the AI what to do. So for now, from what I can see, no one has a decent AI for handicapping. The programmers that are good enough to make one don't, making an AI for say the stock market, would be much more profitable. The people who know handicapping don't make an AI because they lack the programming skills needed. So for now, handicappers are safe from an AI.

IMO, experience and a decent software analytical tool that you are familiar with is a better solution than something new with a catchy buzzword. Handicapping is a brutal game and the more I play, the more I learn. I have never "gone pro" but I would think that the daily grind would be really brutal, which is why I play for fun when I have time.

Having been a professional software developer for over 20 years and having programmed for over 40 years, I can tell you most software is written by people who have mediocre skills at best. The human mind on the other hand has developed thru evolution over millennia thru constant survival testing. Right now, I bet on people.

Belmont123 · 05-31-2024, 04:39 PM

There's an AI based program called Equinedge (EE) developed by a experienced handicapper. It has a some interesting functions but there's no explanation (other than general) of data points used in the algos and you can't change anything within the platform. EE states it uses the best pace algo out there but I'm not convinced based on some of the outputs.

IMO, its a bit light on overall on content and functionality. One still needs to apply good old fashioned capping as it simply looks to identify contenders.

cdax · 05-31-2024, 07:34 PM

Quote:

Originally Posted by Belmont123

There's an AI based program called Equinedge (EE) developed by a experienced handicapper. It has a some interesting functions but there's no explanation (other than general) of data points used in the algos and you can't change anything within the platform. EE states it uses the best pace algo out there but I'm not convinced based on some of the outputs.

IMO, its a bit light on overall on content and functionality. One still needs to apply good old fashioned capping as it simply looks to identify contenders.

I don't think their pace numbers are much better than any other out there. But their "GSR" rating shows a lot of promise. It uses ratings from the horses Dam and Sire and extrapolates a number for it's offspring. Gives me help with first time starters since I don't keep data (or have the mental capacity) for following pedigrees.

Just my 2 cents.

Belmont123 · 06-01-2024, 12:52 AM

Agree, the GSR ratings are useful for the race type you note.

teb21 · 06-01-2024, 11:20 PM

I tried using a neural network long enough ago that it came on a floppy. I hope that many people use AI as there may be value in betting against AI selections.

05-20-2024, 05:03 PM	#1
Old Arkie Gal Grade 1 Join Date: Apr 2008 Posts: 154	Ai I have been away but I haven't seen anything posted about AI and the potential effect on handicapping. I don't know much about AI or its applications but it seems it might have a deleterious effect on it.

05-31-2024, 04:39 PM	#3
Belmont123 AlwNW1X Join Date: May 2024 Location: New York Posts: 5	Ai There's an AI based program called Equinedge (EE) developed by a experienced handicapper. It has a some interesting functions but there's no explanation (other than general) of data points used in the algos and you can't change anything within the platform. EE states it uses the best pace algo out there but I'm not convinced based on some of the outputs. IMO, its a bit light on overall on content and functionality. One still needs to apply good old fashioned capping as it simply looks to identify contenders.

06-01-2024, 12:52 AM	#5
Belmont123 AlwNW1X Join Date: May 2024 Location: New York Posts: 5	Ai Agree, the GSR ratings are useful for the race type you note.

05-21-2024, 12:53 PM	#2
DanBoals Grade 1 Join Date: Apr 2009 Location: Reno, NV Posts: 311	Hi Old Arkie Gal, Are you from Arkansas? I grew up outside of Tulsa, so same area if you are... The only person I know trying to sell an AI system is Dave and he talks about what he has at: https://www.community.horsestreet.com/discussions/p1 From what I see and having talked to him in person last summer about it, it isn't as good as RDSS from my standpoint. AI is really a marketing term. What people are calling AI are actually what programmers have called Expert Systems for well over 30 years. The AI's being touted use what we call Neural Networks, which means they vary constants and weighted variables slightly and see what the impact is on results. Neural Networks have been around for about 30 years and I used them in factory automation and motion control software in the late 90's. Howard even tried them but did not have the technical knowledge to do much with them and so rejected the idea. What is called generative AI, like ChatGPT is basically using an expert system trained thru a neural network that has criteria set by the developers and is focused on conversing with users. When they give results like black nazi soldiers in the german army in WW2, it is a hilarious example of how bad they are. What we call AI will improve and may someday actually be AI. Right now, the problem people have with AI in a field like racing is that the programmers do not understand handicapping enough to tell the AI what to do. So for now, from what I can see, no one has a decent AI for handicapping. The programmers that are good enough to make one don't, making an AI for say the stock market, would be much more profitable. The people who know handicapping don't make an AI because they lack the programming skills needed. So for now, handicappers are safe from an AI. IMO, experience and a decent software analytical tool that you are familiar with is a better solution than something new with a catchy buzzword. Handicapping is a brutal game and the more I play, the more I learn. I have never "gone pro" but I would think that the daily grind would be really brutal, which is why I play for fun when I have time. Having been a professional software developer for over 20 years and having programmed for over 40 years, I can tell you most software is written by people who have mediocre skills at best. The human mind on the other hand has developed thru evolution over millennia thru constant survival testing. Right now, I bet on people.

06-01-2024, 11:20 PM	#6
teb21 Grade 1 Join Date: Oct 2013 Posts: 121	I tried using a neural network long enough ago that it came on a floppy. I hope that many people use AI as there may be value in betting against AI selections.