Why has technology left handicapping in the dust?

[Jeff P]

Originally Posted by Bustin Stones

"It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug. Sarah Connor : Skynet fights back."

Thanks. (Got a laugh out of that.)


-jp

.

[PaceAdvantage]

Originally Posted by o_crunk

This is prolly the best comment in the thread. Some of you are telling on yourselves about your own working knowledge about how “AI” works in horse racing. Might want to let an AI loose on all the posts in the history of this board and I would be happy to have my machines bet against it.

Bottom line is you will spend the majority of your time data cleaning more than any other task. If you think the data that’s freely available (or even exorbitantly priced available) and that is going to be good enough for an “AI” to bet, you are sorely misinformed.

AI technology and hardware have made great strides lately. A totally separate AI application could be created to decide which inputs are optimal.

The possibilities are limitless at the high end (and by high end, I mean those with the means {$$$} and the desire).

[ReplayRandall]

Originally Posted by Jeff P

I would agree with the bolded part. Cleaning the data prior to analysis is huge.

Not so sure about the sentence after that.

The key would be making some very astute observations about horses and races very few players have ever thought of before.

And from there creating custom features from those observations, adding those features to the data - and quietly using them.

I've personally seen it done.

Horseracing pools are different than financial markets in that the more money on a horse in the parimutuel pools the lower the payoff.

Whereas increased money flowing to the shares of XYZ Corp results in a higher share price (at least until big money cashes out.)

Also the total addressable market for horseracing software/algorithms is the tiniest imaginable fraction vs. that of currency trading (Forex), and world stock markets (NYSE, Nasdaq, S&P 500, Stoxx, etc.)

I think that's the real reason why big data innovation hasn't found its way to horse racing.

I also think both CJ and Ralph said it much better than I ever could several pages ago.




-jp

.

Bolded lines are the truth of the matter....You can't buy this, it's not for sale.

[o_crunk]

Originally Posted by PaceAdvantage

AI technology and hardware have made great strides lately. A totally separate AI application could be created to decide which inputs are optimal.

The possibilities are limitless at the high end (and by high end, I mean those with the means {$$$} and the desire).

Ai hAs MaDe GrEaT StRiDeS lAtElY.

No doubt, it can do simple things like "write an ETL program to scrape a website" or "Write a SQL query for X given Y schema". But when you get into the basic math of a given domain, it confidently gets the basics dead wrong. Dead wrong. And the users asking the questions often don't know how wrong it actually is as my link shows. It got one thing right - the implied prob of a 5/2 shot with no takeout. Give it the takeout and it fails miserably and worse a user will be fooled into believing it right. Hell, half the people in this thread probably couldn't tell you it got the implied probability wrong. People are going to pay whatever they are charging a month for this crap thinking they are going to change the world. It's a lot like self driving cars. A big stride is made in the perception of what is possible but in practical reality, it is *very* far away.

I agree with the others that the most powerful stuff is not for sale and never will be. Spinning up your own metrics is the most powerful thing you can do with the data. I also agree, and said earlier in the thread, the amount of paying customers racing has is small and the amount who would pay for an AI is even smaller than that. Like maybe a couple hundred people total.

[thaskalos]

I keep hearing the argument that we can't profit in the game by using information that's readily available to the masses. And yet, I can't help but feel that our main edge as players lies not in our "tools"...but in the creativity with which we use them.

[azeri98]

Originally Posted by MJC922

With any computer-based approach it's only going to be as good as the inputs. If you feed AI DRF data with Beyer figures I don't know how much it can actually learn. There's a lot of relevance to who takes money too, and how does that get presented to AI. Coding for horse racing is incredibly labor intensive; just to get the data in position to be analyzed is a big effort. Then once you have said 'system' you need today's data to be crunched because it needs to look at grouped data not just raw data on a card, everything changes relevance within the group. Computers do an awfully good job if they're properly spoon fed really good inputs, but where are those inputs coming from? Beyer figures? CJ's pace figures? Raw times? My stuff? Nobody has my stuff in CSV except for me. So then they have to build something like I built or steveb's stuff and feed that 'goodness' to the AI. Otherwise it gets nothing and nowhere with just raw PPs IMO.

I disagree, not only will the AI read the data it will be able to read the trip notes and determine all the factors into which horse will win a race. That's what an AI is being built for. To figure out problems in seconds that a human brain can't. Other than cheating it will be able to decipher everything else. Not only that an AI will be able to access any information on the internet. You won't need to enter any data. The AI will be able to access it. All you will need is the track and race that you want to play.

[thaskalos]

If handicapping horses is such a "science"...why are we still waiting for the first profitable "Black Box" handicapping computer program?

[azeri98]

Originally Posted by thaskalos

If handicapping horses is such a "science"...why are we still waiting for the first profitable "Black Box" handicapping computer program?

Because a black box can't decipher the human element like cheating and bad rides by jockey's or if a horse isn't in top shape for the race or doesn't feel like running that day. I don't even think a fully evolved AI will be able to determine the human and animal element from the sport.

[classhandicapper]

IMO, a fertile area for profits these days is the “exceptions” to the rules.

In every area of handicapping there are things that have been repeated over and over in books, articles, social media, and on TV that have essentially become handicapping axioms that almost no one questions. But within these generally true things there are exceptions to the rules that even sharp players can misunderstand.

Whether it’s pace/time, flow, bias, ground loss, trainer, class, etc… , within each factor there are little applications where the axioms are either wrong or misunderstood and you can find horses that you can upgrade/downgrade that other people are taking at face value and others where almost everyone else is upgrading/downgrading them and shouldn’t be.

[dnlgfnk]

Originally Posted by azeri98

I disagree, not only will the AI read the data it will be able to read the trip notes and determine all the factors into which horse will win a race. That's what an AI is being built for. To figure out problems in seconds that a human brain can't. Other than cheating it will be able to decipher everything else. Not only that an AI will be able to access any information on the internet. You won't need to enter any data. The AI will be able to access it. All you will need is the track and race that you want to play.

Just so much science fiction floating around, due to the otherwise great successes in data mining.

This is actually stating that outcomes are predetermined, i.e., determinism, causal inevitability. Irad Ortiz Jr. has no free will. Barbaro will injure himself coming out of the gate. After determining all the multiplicity of factors in the Preakness, Sunday Silence is exactly a nose better than Easy Goer.

Last I checked, correlation still does not imply causation.

https://www.amazon.com/gp/product/01...53750f15dbd60b

[classhandicapper]

Originally Posted by thaskalos

I keep hearing the argument that we can't profit in the game by using information that's readily available to the masses. And yet, I can't help but feel that our main edge as players lies not in our "tools"...but in the creativity with which we use them.

This is closer to what I was saying in my last post.

Let’s take a biased track.

Everyone understands biased tracks conceptually.

Everyone can see when a track is extremely biased.

But not everyone understands that these biases impact each horse differently depending on their individual attributes even when they have the same style, how it can screw it up the speed figure making process depending on how the figure maker interprets the results, how they impact the typical pace/final time relationships etc…

People understand the generalities, but upgrade and downgrade mistakes are made all the time. So if you can understand just a few of those situations better you’ll have an edge.

[PaceAdvantage]

Originally Posted by o_crunk

Ai hAs MaDe GrEaT StRiDeS lAtElY.

No doubt, it can do simple things like "write an ETL program to scrape a website" or "Write a SQL query for X given Y schema". But when you get into the basic math of a given domain, it confidently gets the basics dead wrong. Dead wrong. And the users asking the questions often don't know how wrong it actually is as my link shows. It got one thing right - the implied prob of a 5/2 shot with no takeout. Give it the takeout and it fails miserably and worse a user will be fooled into believing it right. Hell, half the people in this thread probably couldn't tell you it got the implied probability wrong. People are going to pay whatever they are charging a month for this crap thinking they are going to change the world. It's a lot like self driving cars. A big stride is made in the perception of what is possible but in practical reality, it is *very* far away.

I agree with the others that the most powerful stuff is not for sale and never will be. Spinning up your own metrics is the most powerful thing you can do with the data. I also agree, and said earlier in the thread, the amount of paying customers racing has is small and the amount who would pay for an AI is even smaller than that. Like maybe a couple hundred people total.

Wait...you're going to judge all this based on chatGPT or whatever other chat AI is out there?

A few of you sound like the guys on the trading floor back in the late 90s muttering to themselves that "no computer can ever outtrade me!"

Now there is no trading floor anymore...LMAO

[thaskalos]

Originally Posted by classhandicapper

This is closer to what I was saying in my last post.

Let’s take a biased track.

Everyone understands biased tracks conceptually.

Everyone can see when a track is extremely biased.

But not everyone understands that these biases impact each horse differently depending on their individual attributes even when they have the same style, how it can screw it up the speed figure making process depending on how the figure maker interprets the results, how they impact the typical pace/final time relationships etc…

People understand the generalities, but upgrade and downgrade mistakes are made all the time. So if you can understand just a few of those situations better you’ll have an edge.

Think about the speed and pace figures that everybody considers so "commonplace". The common misconception is that no one can make money with the figures, because they are right there in front of us, staring at us from the page. But there is an art to interpreting the speed/pace figures...otherwise, which of the horse's figures are you going to use? What is meant when a handicapper calls a particular horse, the "figure" horse?

Is there really no room left for creativity in the interpretation of these "commonplace" figures? I don't believe that for a second...and the same can be said about the other handicapping factors like "form", and "class".

[thaskalos]

Originally Posted by PaceAdvantage

Wait...you're going to judge all this based on chatGPT or whatever other chat AI is out there?

A few of you sound like the guys on the trading floor back in the late 90s muttering to themselves that "no computer can ever outtrade me!"

Now there is no trading floor anymore...LMAO

Do you feel that there is a commercially available computer program out there that can outhandicap you?

[PaceAdvantage]

Originally Posted by thaskalos

Do you feel that there is a commercially available computer program out there that can outhandicap you?

I'm not looking for that. I'm looking for something that will help me create my own program, so to speak.

I'm not aware of any commercially available black boxes that are documented winners, so it sounds like they don't exist?

But again, this thread was never about creating or selling a black box.