-
Overfitting: The art of avoiding overfitting and the importance of comparing training and test results to detect it.
-
ROI Calculation: Different methods of calculating ROI, including the impact of flat stakes versus stakes based on odds.
-
Feature Selection: Emphasize the importance of selecting the right features and avoiding highly correlated ones.
-
Data Integrity: Ensuring data is correct and free from leakage is crucial for accurate model performance.
-
Practical Advice: Tips for feature reduction, such as using heat maps and regression, and the importance of starting small and scaling up.
Other Suggestions:
-
Feature Selection:The importance of selecting the right features and ensuring data integrity to avoid data leakage.
-
Model Testing: Three-phase testing approach:
-
Initial Test: Use a test set immediately after training to get a high-level idea of model performance.
-
Paper Trading: Use tools like Flumine in paper trading mode to confirm the model's performance.
-
Live Trading: Deploy the model with real money but small stakes to see how it performs in real-world conditions.
-
-
Consistency: Ensure that the process used in live trading is identical to the one used in training and testing, including using the same libraries, data types, and column locations.
-
Starting Small: Begin with a simple model and a few features to get the process right before scaling up.
These suggestions aim to improve model performance and avoid common pitfalls in model training and deployment.