Kellycriterion

[timseymour42]

In this update, I did the following:

• Created a Kelly Criterion class to calculate the percentage of the bankroll a bettor should use
• Revised the NN_Runner and XGB_Runner files to indicate the correct fraction of the bankroll for each bet
• Created a Kelly Criterion test file to ensure the calculations are accurate
• Refactored color logic in the NN_Runner and XGB_Runner files
• Updated the Readme to reflect the changes

[kyleskom]

Will review this soon. One thing I'm thinking though is to have this behind a feature flag.

[fciaf420]

I am very new to this, how can I pull thos code into collab and run it to get the kelly criterion results ?

[timseymour42]

I am very new to this, how can I pull thos code into collab and run it to get the kelly criterion results ?

You can replace the git clone command with this: git clone -b kellycriterion https://github.com/timseymour42/NBA-Machine-Learning-Sports-Betting.git

You can also plug the model's win probability into this calculator: The kelly criterion feature is not a part of the main code base just yet.

[timseymour42]

Will review this soon. One thing I'm thinking though is to have this behind a feature flag.

Yeah, that's a great idea! I can implement this. Before doing so, I just want to be sure that my code would be used because there is another contributor that implemented the kelly criterion feature. So lmk.

[kyleskom]

Will review this soon. One thing I'm thinking though is to have this behind a feature flag.

Yeah, that's a great idea! I can implement this. Before doing so, I just want to be sure that my code would be used because there is another contributor that implemented the kelly criterion feature. So lmk.

Yup this will be used / merged if we can have this behind a feature flag.

[onlyadamscott]

Such good work. Thanks guys! I'm working on a native Sentiment agent. Still testing for accuracy.

[BG2011]

Great addition man ! Small question this will work for the money line but what about the points ?

[timseymour42]

Great addition man ! Small question this will work for the money line but what about the points ?

Currently, the output of the models is a percentage chance that a team will win on the ML. You could raise an issue that you would like to see predictions for the point spread.

[kyleskom]

Tested and verified. Works as expected!

[kaiforsyth]

This breaks when not passing the -odds argument to use custom odds:
TypeError: '>=' not supported between instances of 'str' and 'int'

Need to convert the input to an integer.

[kyleskom]

This breaks when not passing the -odds argument to use custom odds: TypeError: '>=' not supported between instances of 'str' and 'int'

Need to convert the input to an integer.

Can you open an issue.