Custom Loss Function - Machine Learning

Job ID: 34201941

Budget: $10 – $200 USD

I am looking to develop a custom loss function. Essentially, this is for sports betting. Typically, when I run models building it on my own or running it on azure auto ml, the custom loss functions are tying to maximize accuracy (# of correct predictions). However, that isn't my goal. My goal is to maximize profit. As some samples have better odds, accuracy isn't important to me. I could have s 50% accuracy, but have $2,000 in profit. Likewise, I could have 65% accuracy, and have -$1000 in a loss.

Each sample has odds in decimal format. So the profit model would be:

If correct : Total Profit = odds + profit | $11.50 = $1.5 + $10
if incorrect: Total Profit = Profit - 1 | $8.5 = $10-1.5

So, the goal is to maximize profit, as each previous results profit would be carried to the next sample.