Strategy: Multiclass

Multiclass trading adds a 3rd category ("class") in addition to the usual up and down predictions The problem with simply predicting up or down is that, for example, with $100 stock, if the AI predicts it will go up only 1 cent for the day, that seems not much better than random. And even if the prediction is correct, it would make less than the trading costs. To solve this, we divide the training data into up, down, and mid. "Mid" is for all the trades where the price did not move much. The AI model can then learn what technical indicators cause the stock to make a big move, which is what we want to focus on. We then sit out any trades where it predicts there will be only a small movement ("mid"). In day trading, transaction fees are one of the biggest reasons people lose money, so having a strategy that trades more selectively can significantly help with that.

To test this theory, we ran a backtest, which means we simulated how much profit it would have made if we had traded it for real using historical data. This is a good indication of how well the Multiclass strategy would do in live trading for real.

Results: In our backtest, the AI model's accuracy (up or down prediction) was ______.
The profit/loss backtest had a Profit Factor (PF) of ____ and is shown in the table and chart below. A PF greater than 1 means it would have made a profit. This backtest does not include commissions or the bid/ask spread.




You can run this backtest yourself in our Multiclass Google Colab notebook: Pickstocks.com - Machine Learning. It uses our open-source GitHub repo at Pickstocks GitHub. You can also change the data or AI strategy yourself if you want and then backtest it using our repo.