Here comes the AI

[Dave Schwartz]

Interesting AI article.

Betting On Horses with No-Code AI

[JustRalph]

Sept of 2021 article.

Any follow up?


“Horse racing gambling is basically the suckers against the quants," said Rossi. "And the quants are kicking the [crap] out of the suckers."

At least they’re honest

[Robert Fischer]

smart guys can scour the pools for inneficiencies.

Some very likely know how to see further into data such as coverage multi horse races (gives them an odds model of later races including which FTS are taking money, gives them an unfair advantage in wagers like the jackpot single ticket coverage)

we can easily do three races/placings, even though regular players are only given two races/placings.

criminal teams can use inside information to know things that others can not know. This includes which horses will not receive the undetectable performance enhancers that have flooded the game. They also create their own luck from things like having some influence over the top riders in a colony and occasionally knowing that a favorite/2ndChoice will not hit the exacta.


the rare occasions when I beat the game are an extreme ego boost.

I know exactly how the quant types think (not all... some are genuinely good people, and then some are schmucks but they understand that intelligent or simply masters of the craft exist in the world and could possibly cross their paths) when it comes to superiority in themselves and disbelief in myself (or a less interesting generic population of non-quants and non-insiders)(borr-ing if not personal... I got Jay Z blasting in the earphones let me get mad. Get mad you sonnofabitch get mad PUT THAT COF-FEE DOWNN).
"You can't do that"

They don't believe it.

sometimes the dickheads think you can't beat THEM sometimes those guys or regular losers think you can't beat THE GAME.

horse racing sucks most of the time with opportunities. Takeout needs to be reduced. And the same people (you and i) who whine about being thought of as a sucker, we want takeout reduced and the marketing improved.

but it sure is beautful on occasion.

speaking of the crappy marketing

Twitter

I know twitter can be some faggots, but I expected a few people to like or laugh out loud or whatever to that good tweet.

makes me want to crush a few quants and fuck up a few normies on my sunday watch list horses



Then you have all kinds of suckers.

suckers who think they can beat this game who personally can't beat it.

suckers who think no one can beat it.

suckers who think the celebrity masters of the game in the game want to discuss the game with street-level masters of the game.

suckers who think the quants can't have better pool data to be able to see things like uncovered jackpot wager horses or pick 3(or further??) probables

suckers who think the quants wouldn't attempt to build it

suckers who don't believe star horses are being injected with undetectable forms of EPO and other drugs

suckers who think an 'overage' of a detectable race day medication is the PED, or that such overages are the barometer of cheating, or that a suspension for such overages indicates cheater

suckers that think that herd behavior by them is going to lead to success.

suckers who think the trainer who LARPS as a villain and has some overages probably cheats but that the genial old man who exudes class and trains for wealthy owners would never cheat

or perhaps a guy who specializes in the turf and is built up to a veritable cult of personality is just 'good' and classy and would never cheat like the villains.

or that supertrainers who use uncoupled entries and the top few jockeys of circuit, would never use that power as influence

or that they would never partner with teams

or that top jockeys would 'never risk it'

I don't know shit about most things aside from Horse Racing, Baseball, and Basketball. Learning about horse racing makes me wonder if the rest of the world is as wild as horseracing.

We have a ponzi scheme of cheating. It's not sustainable. There's a limit to how many cheaters we can find to support the top level cheaters. Most people would never cheat. I can't see it happening.

__________________
Preparation. Discipline. Patience. Decisiveness.

---

Last edited by Robert Fischer; 04-02-2023 at 09:38 AM.

[Jeff P]

From the article:

All Akkio needed was data to run through its prediction engine. Enter Chris Rossi. He's the horse betting expert who...

Who among other things, actually posts here once in a while.

Interesting that the model trained on a small dataset.

700 rows and a limited number of columns.

Also interesting that the model was successful in forward testing (for two race days) after it was developed.

Proof of concept I guess.




-jp

.

[vegasone]

There is something about short term results that seems to fool people. Lets see how it does over tens of thousands of races.

[classhandicapper]

I don’t know anything about how AI works, but it seems similar to running PP data through an Excel regression, getting weights for each factor, using the weights to make selections going forward, adding new data as races are run, and rerunning the regression to get updated weights for each factor and so on.

If you were doing this with Excel, you might then break it up by dirt/turf, sprint/route, high class races and low class races and so on until you had a different model for each type of race.

That’s sort of the Excel project I started this year that I’ve posted data on where I was getting 36% winners and outperforming the track take significantly in dirt stakes over 8 1/2 years. That was with pretty basic data and not even broken up by sprint/route yet.

The difference being that AI is doing all the work and learning behind the scenes. That’s obviously one hell of a lot easier and will probably work better on raw data.

[Dave Schwartz]

Originally Posted by classhandicapper

I don’t know anything about how AI works, but it seems similar to running PP data through an Excel regression, getting weights for each factor, using the weights to make selections going forward, adding new data as races are run, and rerunning the regression to get updated weights for each factor and so on.

If you were doing this with Excel, you might then break it up by dirt/turf, sprint/route, high class races and low class races and so on until you had a different model for each type of race.

That’s sort of the Excel project I started this year that I’ve posted data on where I was getting 36% winners and outperforming the track take significantly in dirt stakes over 8 1/2 years. That was with pretty basic data and not even broken up by sprint/route yet.

The difference being that AI is doing all the work and learning behind the scenes. That’s obviously one hell of a lot easier and will probably work better on raw data.

What you have described is not even close to the capabilities of a robust, deep-learning/deep understanding AI engine.

What you have described is going to be purely linear.

That is, the best horse, is always the best horse.

To my knowledge, all the whale teams use the same approach, which is much as you describe.

Not necessarily the precise tools but some form of iterative regression

Chat.GPT v4
me: Please explain iterative regression
then...
me: Pease simplify your answer for 8th grade level

GPT4: Iterative regression means repeatedly adjusting a mathematical model to better fit a set of data. This is done by using the model to make predictions, comparing the predictions to the actual data, and then adjusting the model accordingly. This process is repeated until the model fits the data well enough, based on certain rules. Iterative regression can help improve the accuracy of the model, especially when the data is complex or there are unusual points, but it can also take a long time to run.

To be clear, iterative regression WORKS WELL for the whales.

If you begin now you will (essentially) be doing what they are doing.

But lest one gets the idea that, since the answer for them was/is found in ItReg Analysis, remember that they have a 15+ year head start.

IOW, if you do what they are doing but are not doing it as well, you will be eaten alive.

A sophisticated AI engine (which is NOT what's described in the article) would be able to say, "In this race we bet the best bet because it's the best bet. But in the next race, we bet the worst bet because the best bet will be hammered down to the -3% level by the whales."

IOW, a deep learning system is not linear.

In fact, it is capable of understanding that there are times for linearity, but other times you need anti-linear, reverse linear, or even chaotic randomness.

[headhawg]

"AI" is becoming a catch-all term for things like machine learning or deep learning. The comparison to regression is similar to comparing a Model T and a C8 Corvette. They are both cars that will provide transportation but the level of performance and technology are quite a bit different.

Even within regression there are multiple types. I remember reading on PA a long, long time ago from a member who was quite knowledgeable about statistical analysis. I remember that he posted something about the limitations of multiple regression and that it didn't work in a handicapping environment. I assume he was talking about a least squares method which is what Excel uses by default. I think improved results can be found by using logistic regression; Dave Schwartz might have made some posts about that a while ago. (Or maybe that was about Bayesian stats, I don't remember.)

Anyway, I think learning about AI is worth the effort. I'm reading some books now and plan to enroll in one or two Coursera courses about machine learning. Not sure I will do anything with it but this stuff is good for the brain regardless.

[Dave Schwartz]

"AI" is becoming a catch-all term for things like machine learning or deep learning. The comparison to regression is similar to comparing a Model T and a C8 Corvette. They are both cars that will provide transportation but the level of performance and technology are quite a bit different.

It would seem that you and I are on the same page.

[PaceAdvantage]

When it comes to making money (in any endeavor, really), it's always a good thing to be on the bleeding edge....no matter what "traditionalists" might tell you...

"You can't win with a computer"

Remember that stupid phrase from 20 years ago on this board? LMAO

[classhandicapper]

This stuff sounds amazing.

I've been studying this game for close to 50 years now and very seriously the two times I've had an active database. I've replaced a lot of trial and error, theory, and book learned info with hard data, new metrics, testing, and facts. But it would probably take me another 50 years to learn what one of these new self learning systems can learn and refine quickly.

I don't think it's hopeless though. I still see fertile areas of value when human judgment has to be involved. I think even the best individuals and teams miss some things and make the occasional mistake.

[classhandicapper]

"But lest one gets the idea that, since the answer for them was/is found in ItReg Analysis, remember that they have a 15+ year head start.

IOW, if you do what they are doing but are not doing it as well, you will be eaten alive."

I have no desire to compete head to head with math geniuses at building a math and statistics based model.

I love the old joke, "How do you beat Bobby Fischer?"

Ans: Play him at something other than chess.

For gambling purposes, I'm trying to focus on things that require some human judgment. For me, bias is easily the most fertile ground. You can automate bias analysis to a point (and I have), but you still have to watch a ton of replays to get it right more consistently. Even most serious players will only do that for 1 or at most 2 tracks. So occasionally things fall through the cracks.

[Scamper]

Horse racing is a business, not a sport.
AI would be effective at finding business
trends much better than it would at
handicapping the ponies. IMO

[Dave Schwartz]

I don't think it's hopeless though. I still see fertile areas of value when human judgment has to be involved. I think even the best individuals and teams miss some things and make the occasional mistake.

I agree with you!

But, this presupposes that one's human judgement can outperform the robust statistical gyrations of the whales.

Also key is to understand what areas the whales' approach are likely to overlook.

For the record, it isn't really about mistakes.

IOW, what can you be expert at that the whales have no interest in becoming experts at.

[thaskalos]

Originally Posted by Scamper

Horse racing is a business, not a sport.
AI would be effective at finding business
trends much better than it would at
handicapping the ponies. IMO

The ancient game of Go is infinitely more complicated than "handicapping the ponies". And yet...a computer was programmed to learn the game by playing against itself for a couple of weeks...and then the computer practically destroyed Go's greatest player, who was considered UNBEATABLE up to that point. In fact...even the computer's programmers were shocked when they saw how easily the Go champion was defeated.

Gentlemen...we have met the enemy...and we are theirs.