Profits with Speculation – Horse Racing AI Model Journey of Money Management

Profits with Speculation – Horse Racing AI Model Journey of Money Management

Greetings! First, some disclaimers:

1. This is for entertainment purposes, also. Yes, also.

2. This is more about discipline and organisation, than it is any ‘methods’. If anything, the underlying ‘method’ here is leveraging Paypal’s working capital program.

3. Ironically, despite #2, expect a rather unstructured approach. So, this may not be for everyone.

About me

If anyone asks; I am a writer. I make my primary income online running a publishing company. It’s just an EIN#, I’m the one man business model (no copyright infringement to that youtube guy, ‘Dan Kloe, Dan Koe?). The key is to write about what you like. I happen to like extremely unique and abstract, yet complex topics and subjects to write about. So, I write for about 5 different platforms, each at different rates, some pay extraordinarily well, whilst others pay peanuts. Well, that’s IM.

So I write, and speak about things, now as writing has extended to ‘content creation’. Think writing meets cameo. That’s what I do. So, here are each of my main streams of IM income. That is, simply, writing. No secret formula there.

Title: Special Advisor
Vendor #1 [ payment: direct paypal – 70% profit after 30-33% commission paid to vendor. Most made in one day = $350]. This one, I’ve been doing for 10+ years! At first, as just a free time extra way to make ‘fun’ money online. You know, keyword “Make money online”. Again, my first intro to this was Commission Junction and my favourite no-longer-existing Copeac. Anyway, I digress. Onto the streams of income:

Title: Engagement Specialist
Vendor #2 [payment: paypal / weekly – lowest payout, at 35% after commission paid, but consistent clients] Most made in one day = $75

Title: Engagement Advisor
Vendor #3 [payment: paypal / 3-5 days after completed work – 60% profit after commission paid, with consistent clients of moderate to high disposable income] Most made in one day = $300

Title: Care Counsler
Vendor #4 and 5 [payment: paypal / direct deposit 3-5 days after completed work – rates vary, but weekly income is just about $200/wk with some sense of consistent effort per 7 days.


Other:
I make and sold prompts on promptbase and other sites. Meh. No interest. I’d rather enjoy what CHATGPT, like this, for pleasure. Which means, I really don’t have any interest staring a ‘tipping’ service, or anything related.



BACKGROUND
So, I started off way back after discovering auto-blogging, and cashcrate. Nothing like setting up an autoblog to get cashcrate, bankrollbucks, treasuretrooper and other GPT referrals of those days.

Then I had a rather short, but successful run in Affiliate marketing promoting streams around PPVs and credit report cpa offers to apartment seekers on craiglist. Copeac was where I connected with my affiliate manager, during lunch, whilst on my 9-5.

Next, I settled on kindle publishing. It was quite easy to run AMS marketing keywords and run and awesome roi.

I was making a full time income with IM. Finally. But, I enjoyed the (false) security of the 9-5. Not to mention it was in the Music Industry, so it was ‘fun’. Finally, I went all in to IM with the focus on the above, since the latter period of 2022. So, it’s been full time since.

But, the only job I’ve had, as an adult, is in the music industry. Corporate at time, but, no structure. Party time. Etc. So my level of structure and disciplined is scattered. So, that’s part of my reason for this.


HORSERACING AND CHATGPT – Reasons For This Journey

1. Discipline. I know, very strange. So, Gambling is the answer? No. And disclaimer – I don’t promote gambline.

So my no-overhead content vendors/clients provide me with consistent work, granted I just make myself available. I left the music industry, but it didn’t leave me. So, instead of being really productive, I’ll just stop working for a few weeks or months after a productive run. Yeah. Not cool.

2. Research – What keeps me ‘available’ and away from these mental ‘benders’ of sorts, is research and speculation analysis. So, when ChatGPT emerged, my work there was some SEO stuff and beating ai content detectors. Then it just expanded to what we’ll be reading about here. More of just f***ing around with CHATGPT – but in a somewhat product to-my-life-way. I'm, essentially a geek, so a night of research is bliss.

3. Logistics – The more I research, the more information I have, and the more I am available. I know, very strange, but it works for me, and, secondly I enjoy writing journals and diaries – just never did it here, yet. So, the long winded point, is that the longer I stay typing about these fun AI adventures, I make more money. Apparently. Based on the data.

Ok, enough. Let’s jump right in with the race analyses.

I. Horse Racing and AI

We are going to ‘build’, or ‘prompt’, GPT into a Feature Engineering Model to predict horse finishers. I’m calling it a Horse Racing Simulator.

Now, this has been studied and researched for years (more or on this in the next section), so, it has been done successfully at varying degrees of success etc. So, it’s not question of IF it can work, it’s a question of HOW , and how to do so in a way that’s strategic and maximizes ROI.


Like I alluded to, I have this uncanny innate desire and skill of research. And, now I have the ‘time’ to do so, and apply it, and so, in my previous "experiments" I see, with focus, this "can work" for ANYONE.

So, here’s our angle: MAIDEN RACES. They are better because odds are higher and payouts are greater. For those that don’t know the SPECIFICS of horse racing, but want explanation, let me know. I’ll write something up, if desired.

So here, the point is not so much of the frequency of accuracy, which at 100% would still yield a small loss if one bet on all races.

The point, instead, is to target specific races, with worthy odds to make a decent profit if even we wager up to $20, during one session.

Next – ‘creating the feature engineering model of CHATGPT (4o)

Here, we have our custom metrics, based on what's been gathered from various discussions, online, over the last 15+ years.
This is cool, because some of the "old internet" data is there, where people really developed some OUTSTANDING handicapping methods.

III. Rules

1. We will bet on no more than 3 races per day, but no more than one race, per track, per day.

2. Max ticket/investment - $20

3. Have fun and enjoy the races.

4. The idea is to utilize this as further investments into 1) speculation (high volatility stocks), 2)Assistants (ChatGPT seems to have that market 3) Marketing

5. Remember, there are unlimited way to make money online.

Now, it’s off to the races!

Today's Races:

Date wise, this will be some day(s) behind.
So, if you need me to explain anything, or if I've missed anything do let me know. I'll let the screenshots speak for themselves:




Continued....

That's interesting...a unique journey, thank you for sharing it with us

Our model seemed to be pretty confidant about this race finish, even after running a 5 minutes to post analyses to catch the movement of odds.
My Bets





The Result: Note. Our model had the EXACT finish. However, horse #1 was disqualified from the 2nd spot. So, we missed the trifecta. But we did get the exacta from that disqualification, one of our back-up bets, which resulted, most likely in a similar payout since this was a $3 wager.

Let's keep this train moving. Amazing. We can travel to any racetrack at ANY TIME. Next stop - WOODBINE - RACE 6!



Let's go straight to the FINAL 5 MINUTE ANALYSIS:



Ok, so based on the movements of the odds, here's our top 3:




Well then. That Ontario sure does look enticing. Our unique GPT model seems to be confident on, at least CATCH THE DRIFT.
So, I'll make a few small bets around those horses.
Remember, the ODDS are pretty good here. So $2, here, $3 here, could return $20+
That's why we have the max ticket at $20. Anyway, our choices, and result......

Our Maiden Simulator GPT Superhero had it right. Somebody knew that Ontario would be competitive. Here's another amazing fact. This is this horse's debut:

The actual winning combo:

These races were done over two days.
Over those two days, due to my concentration on, well - just being online, here's the income from the other endeavours, done simultaneously.

Publishing/Content:

Title: Special Advisor
Vendor #1
6 July - $50
7 July - $465.05

Title: Engagement Specialist
Vendor #2
7 July - $50.34

Title: Engagement Advisor
Vendor #3
6th/7th July - $30

Title: Care Counsler
Vendors #4 and 5
6th/7th July - $30

Already, this "schedule" has yielded strong benefits. Plus, it's hella fun watching the races unfold as I take my "breaks" from writing....
PS: Whilst making these posts, (8th July) I did have an engagement for the Engagement Advisor gig. That 10 minute session paid $26.03.

i still dont understand of your prompt, could you please tell me more?

bosfigur said:

i still dont understand of your prompt, could you please tell me more?

Click to expand...

This prompt is designed to predict horse race finishes, along with providing betting strategy suggestions, based on the analysis.

On average, what is the successful rate of this speculation AI model?

ChocolatePie said:

On average, what is the successful rate of this speculation AI model?

Click to expand...

With maiden races 9 runners or less 70% in getting the top 4 correct.
Order isn't always right, however PLACE seems to be incredibly accurate, which really helps with exactas, which is, ideally, my focus.

I like to keep it in between 5 and 9 runners, preferrably "7". That's the sweet spot where the race order is more often than not, completely accurate. (with the top 3)

This will only work, really for Maiden OR races with long odds and less racing history.
Because the features being weighed are different than just the 'typical form analysis, direct form/collateral form lines)' that most are doing, by defautl, of course.

So, of course, naturally, we’ve got to include losses, as well as wins, too, right?

So, let’s cover the last couple of days, then, onto the next win to show how this amazing tool can, with some focus, and discipline, yield, cool results. Of course, the key is to keep the pressure off, and the emotion out of it, which is, really, ultimately, the other key.

8th July – ROI: -$12 (Delaware Park)

Feedback: Here, I slipped into the typical GAMBLER. Trying to ‘cover all of the bases’. GPT got the win correct, but I played to the underdogs, as a result of LOOKING AT THE POTENTIAL LARGE PAYOUTS. So, in other words, I turned into the gambler. Now, instead of trying to “make up the loss – we just move on. Tomorrow’s another day.


9th July

Approach: No longer will we aim to ‘cover the bases’. To FULLY take advantage of all of the things that we say are advantageous to Maiden (newcomers/debuting horses, young horses), we’re now going to limit EACH WAGER PER TRACK TO $2-$5.

RACE 1 Finger Lakes – Wager/Investment= -$2 – I bet on one exacta that didn’t make it. I actually wanted to get another bet in, but couldn’t do so in time, before post time. However, I NEVER rush to make bets. I heard a quote that said “Don’t worry about missing races/bets”, or something along those lines. It’s better to not bet, then to place a frantic/rushed/emotional bet.

RACE 2 – Finger Lakes again – Wager/Investment - $5. Did an a couple of exacta boxes and a trifecta, which, didn’t hit. Now we are down $7 for the day, and down $19 for the last two days. OH NOES!

Approach: Continue to keep the bets small. Don’t worry about “making up” for the losses. These are Maiden races, perfectly made for GPT analysis, in races where the FAVOURITE doesn’t always in, because, well, so many variables, which we trained our GPT to analyze.

We're not done with 9th July....

It should be noted that prior to this project, my knowledge of Horse Racing was limited to a board game that I had got as a gift when I was about 10. Yeah, a board game. So mind you - I had no idea what "maiden", "pace", "barriers", or any of this Horse race analysis jargon that's part of this, one of the world's oldest means of gambling speculation.

Remember this GPT PROMPT is being used as a TOOL OF ANALYSIS. I am NOT using it simply as a 'do-it-all-spit-out-the-exact-order'
That would be rediculous. I've prompted the GPT to give me actual and practical feedback so that I can make the decision(s).

So, as a result of the various as analyses, I've come to a somewhat (over)generlisation - can be overused at times, but with Maiden races, again, the ADVANTAGE TO US is that uncertainty of the outsiders and unknowns.

Our next target: (9 July)


You can't see it here, but that #8 horse, PEACEFUL WARRIOR has some high paying 15/1 odds.
That #2 horse, All Roads To Denis has reculously LOW odds. Pays low and seems to be the VERY STRONG FAVOURITE TO WIN.

Let's look at some key GPT Analyses, here.



So, we've got that FAVOURITE to win, but this Peaceful Warrior looks awfuly enticing. He's been great in previous races. Improved in workouts. Pedigree looks top, and the trainer stands out here.
Hmmm. I still need more.....



Check out those factors.
1. Market movements - GPT identified this horse as AN OUTSIDER. And predicted TOP 4. Hmmm.
2. My analysis, conclusion and bets(s) - So, here's what I think. Either that favourite horse All Roads Lead to... is going to win, with the 'next favourite', or he's going to get his a$$ smoked by this PEACEFUL WARRIOR. I'm going with the latter.

So, we'll make, again - small wagers, hoping for a big return. If we lose, it's just another few bucks. If we win, well.....
Let's take a look, shall we.
With 2 MINUTES TO POST, we are going to run another analyses, as the final odds money comes in. This is where the insiders make their bets:




Ok.
Based on this, we are going to make a $2 wager/investment into the "safe bet" -
Horse #2, then Horse #3.

I still think PEACEFUL WARRIOR, is not going to be peaceful and is going to smoke both of them. Horse #2, will have to settle for second, me thinks.

Let's check out the results.



We nailed it!

Peaceful Warrior smoked them all.
We placed a $2 ticket on the 'safe bet', and a $2 exacta ticket on our top 2.
Wager / Investment = -$4
ROI = +$70.04

Time to walk away from the RaceTrack now, until next time.
Thank you for your analysis, GPT. Until next time.

Quick update, since I ran 3 races today.
Stakes: $21 out (races 1 and 2)
Returns: $31 in (race 3 win exacta win, from $2 exacta [TOP 2 HORSES PICK/FINISH ORDER] wager/investment)
Up $10 today. Time to walk away from the track. (i.e close the app)

Lessons: In each case, my "second" bet was the winner. I then proceeded for some extra bets (superfectas). Working on that little item of discipline that, again, stops at the pre 5 dollar ticket, UNLESS incredibly confident, which - is really just heightened emotion, looking at the potential dollar amount.

The GPT, based on our training (see first post) delivers strong insights in it's actual description(s) of the analyses, which seems to be consistent to projected performances. My goal each day is to simply 'be up'. Doesn't have to be a whopper each day - just being positive is the top goal.




So, I placed exacta bets on 3/4 and 4/3. Then proceed to add $5 more in superfectas. Why, that darn emotion.
The $2 bet on 4/3 returned $31.80



Other updates: In first post, I mentioned that I am using Paypal Working Capital. I am 57% done with payment, thus far. Initial date of capital was 23rd May. So roughly a month and a half ago.

I think we're onto something here....

Hello all.
Glad some of you find this interesting. A small update.
Like I said, this would be niche, small, unique, and somewhat unstructured; but for me, this helps with data and avoiding repetition of the same mistakes, not yet eliminated during my testing periods a couple of months ago. Some notes:

11 Jul saw a $10 drop into a hole, but then I got a $10 racing credit promo.
Today, I'm up $2, after blowing a $12 ticket, but then winning a $25 exacta, and a $5 place.
I'm up about 75% since beginning this journey.

I'll share screenshots, perhaps with the next race update.

Updated Approach:
1. My $0 up from yesterday, was actually me deciding to take that $10 credit to bet on some SUPERHIGHFIVES. It was also on an CLAIMING race and not a MAIDEN. Although I had a "general" model before, this has given me an idea to, now, create a Prediction Model (Prompt) specifically for the various classes (i.e. MAIDEN, CLAIMING). However, again, MAIDEN seems to offer the best odds.

2. Bets on the day before - Because this prediction model/prompt is accurate, or - at its very worst, helps me make educated decisions, I'm going to get organised with this, as this diary, as I mentioned, really helps me organise this, to, eventually grow it. This is, uniquely (or oddly) enough, is very therapeautic to my research mind. Before aol (or sportsbetting) was a thing, I played online casino video games excessively. Never cared for actually walking into a Casino besides just a curiosity. Once I actually did I played a few slots, won $122, and walked around Windsor Ontario until about 4am until returning to my hotel. So, there's a very specific muscle that's being excercised here, for all you neurologists, where it's natural for me to treat this more as analysis. Using CHATGPT in this way, I know, will lead to this little 'recreational project' turn into something more - another reason for me doing this journey.

That's my long about background info way to say that I'm going to make bets the day before. Or, at least run the reports. Since these are MAIDEN RACES (young/never won before/debuts) , the FAVOURITE is often not, the one to win. So, I noticed that the FINAL 2 MINUTE RESET analyses, usually doesn't change the finish THAT MUCH. So, my new approach will to do the main analyses the day before (or prior, even two days!), then start with supplemental reports 30 minutes to post, 15 minutes to post etc.

I have a more refined betting method -
EXACT BOX, followed by WIN/PLACE/SHOW bets as suggested by our GPT Horse Racing Simulator Prompt (just a reminder to all, that this is just a prompt, specifically turned into a Feature Engineering Model to predict outcomes based on data fed to the GPT4o)
Again, this works for MAIDEN races due to the long odds and unknowns (horses making their debuts, or just a few starts in their 'career').

This avoids trying to 'cover all the bases', emotionally, thus inflating the ticket. I'm up about 75%-78% right now, with one miss on a non-maiden race. The first race of the day, 'THE GAMBLER' is all pumped up and ready to spread all the money all of the place.

3. The idea will be to spread more bets across many races. Eventually, of course, the bet amount will be increased, but, for now, this seems to be working.

To the Horse Enthusiasts.
I may not get all the terms properly, so please forgive and feel free to correct.

To the Computer Programmers.
I think ChatGPT is pretty cool and I have no previous coding knowledge, outside of typing "home" on one of those really old computers. I had no idea what Python was prior to ChatGPT (still don't, really) and just heard of Feature Engineering a few months ago. So, same - excuse any irregular usage of any of those terms.

To all
See you at the races.

Hello folks.
No worries, at all. I have not abandoned this.
For pete's sake, look at my begjn date. Surely, I am not one to start 'just another' abandoned journey

So, I decided to cash out at 95% profit. I kept fluctuating, so, as one of my fellow friends named, (Orin, after a star, no less,,says, "Don't be afraid to take profits."

So , I take profits at about 95ish% percent profit.

To follow up above, I've galloped (no pun) on my Paypal capital repay at 80% now. Almost there, to the breakthrough.

Ok. As you were.

Great journey (and it's also quite unique). You will need a lot of luck so best of luck OP

following this, i am not really into horse racing but it would be fun to see your breakthrough.