Machine Learning
 

Wondering if anyone has delved into Machine Learning for racing?
Was it worthwhile? Did you gain any advantage.
I don't expect anyone to give away any secrets, but a simple yes or no to both questions might be helpful :)

CP,

Yes I have, Yes it is profitable.

Most people don't have the required data to do it properly, or the right skills to do it properly.

If you are going to use ML you are going to have a lot of data, understand ML at least a little bit (do a few youtube 24 crash courses on it) and be able to program in python.

Hope that helps.

I was expecting your input, although my python has progressed i still haven't used it for data collecting.

It just takes some time and a whole bunch of commitment, then some good amounts of luck.

The last 20 years has been great for people who can code and applied it to a hobby. That will remain for another 5-10 years and then coding will be mainstream meaning that being able to code will be less and less of an advantage.

At that point it will be about data analysis and who can do it best. Which is why its important to learn about Machine Learning.

I have been doing work on this recently, I have many years of greyhound data from my Surepick business, particularly valuable is our proprietary info.


I use RStudio for the ML analysis. As Shaun notes, the main issue is understanding ML, statistics and what combinations of your data can be best used for predictions. It is critical to remove as much co-linearity as possible. this is where the trial and error comes in, looking for the most significant data when creating Training and Test sets.

To get into ML, I found the kaggle ML resource fantastic. You can build your skills working on their Titanic ML example, where you try to predict which passengers live or die based upon specific data factors.



Grant

Sorry for the delay in replying, I do appreciate everyone who has provided input. Given me the incentive to at least give it a try.