• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

Talking to your CSV on your laptop

MySportsAI

MySportsAI

Filly
A good judge I was reading the other day stated that like all big tech innovations LLM's will eventually migrate to your local machine. You will talk to your AI LLM locally on your machine because thats where it will reside. Couple this with an LLM trained for your specific use eg betting and it has a certain appeal. We can already install the required libraries to talk to our CSV files locally in natural language. I gave this a try this morning and here is what happened. I first issued the following prompt on a file of flat data for 2016 to 2017

What is the average pdsBtn for all races

Note the use of the exact column name pdsBtn for pounds beaten

The LLM took 3 minutes to answer and gave a correct or at least it looks correct. It also spits out the command it is going to use to get the answer so I am con fident the answer is correct

Next I asked it

What is the average of the pdsBtn column for all races of raceType Handicap

it took about the same length of time but it also got it wrong in that for some reason it examined the column raceTypeLTO (race type last time out) despite me using the column name raceType explicitly

If you are interested in trying this yourself let me know and I will walk through the setup
 
MySportsAI said:
Changing the prompt to

What is the average of the pdsBtn column for all races of raceType = Handicap

got it to work OK
Click to expand...
What do you think about diminishing edge because even total neanderthals like me can now build complex rating and betting models, anyone can do it, will that mean less advanced methods will become more profitable.
 
Its a fair point, I think the models will become more sophisticated with input data. For example some syndicates are not just feeding last minute odds into their models but also paddock inspections
 
The issue was that my GitHub Copilot subscription became unusable due to their plan changes, and the monthly limit will expire within two days. However, as I mentioned before, my main objective wasn't tied to the tool itself; the goal was to build the overall infrastructure for data analysis, strategy building, and testing, and that is mostly complete. My Reddit post best describes this: I successfully managed to integrate AI into every point of the development loop.
 
You must log in or register to reply here.
Back
Top