Introduction to Trading Strategy and Indicator Update
Hello STK,
I have made significant updates to the code, replacing the old Telegram system with a more advanced and efficient : NET 6.0. This improvement eliminates the need to grant FullAccess permissions, a breakthrough achieved thanks to ClickAlgo's valuable guidance and technical support. The messaging system is now more robust and secure. The next step is to carry out tests on MAC OS to ensure the full functionality and reliability of our product, which has already been tested on Windows 10.
Strategic Considerations for Trading Profitability
Volatility and Market Movement: Recognizing the critical role of high volatility is essential for successful trading. It acts as the primary driver for substantial price movements.
Long-term Trend Analysis: Utilizing tools like the Extended Period Moving Average offers insights into market trends, while the MACD (Moving Average Convergence Divergence) enhances trade initiation by aligning trend, momentum, volatility, and the predictive insights from machine learning models.
Key Trading Concepts
Risk/Reward Ratios and Accuracy: An essential aspect of trading strategy is understanding the balance between risk and potential reward. This balance is quantifiable through risk/reward ratios, which should ideally align with the accuracy of the underlying model. Strategies with high hit rates might have less favorable risk/reward ratios, whereas strategies with lower hit rates can afford more advantageous ratios, given the model's accuracy.
- Recommended Risk/Reward Ratios Based on Model Accuracy:
- For models with an accuracy of 50%-60%, a risk/reward ratio of at least 1:2 is advisable to compensate for the lower hit rate.
- With accuracy levels between 60%-70%, a more balanced ratio, such as 1:1.5, can be efficient.
- Models achieving over 70% accuracy can operate effectively with a risk/reward ratio closer to 1:1, reflecting higher confidence in the trading outcomes.
Enhancing Accuracy through Indicator Confluence: Waiting for the convergence of indicators can significantly improve the underlying model's accuracy. This strategic patience allows for more precise entry points, optimizing the risk/reward ratio by leveraging the collective predictive power of multiple indicators.
Transaction Costs: Consider the impact of commissions and spreads on the required hit rate for profitability.
Capital Management: Practicing effective risk management, such as limiting exposure per trade, is critical for long-term profitability.
Technical Indicator Update
- Latest Developments: We've updated our indicator to Telegram NET 6.0, improving its functionality and ensuring compatibility across platforms. Preliminary tests on Windows indicate stability, with ongoing verification on Mac to ensure optimal performance.
Model Testing and Optimization
Accuracy Focus: It is crucial to focus exclusively on model accuracy during the testing phase. I chose to allocate 98% of the data to training and only 2% to testing because machine learning models tend to lose accuracy quickly with new data, beyond what is observed in the test or validation sets. For this reason, it is advisable to restart the indicator periodically to update the model, thus mitigating the risk of over-fitting to historical data, decreasing temporal autocorrelation in financial markets and inherent volatility.
Indicator Period Adjustments: Modifying the indicator periods can boost the model's hit rate. Your feedback on further improvements is highly appreciated.
Seeking Optimal Test Accuracy
Balanced Accuracy in Testing and Training: Achieving high accuracy in both phases is key. Fine-tuning the k parameter of the kNN model can improve test performance due to the varying data volume between training and test sets.
Optimal Data Volume for Testing: I recommend a test data volume of 50 to 200 points for robust accuracy metrics, ensuring the model's predictive accuracy on new data is not compromised.
Additional Resources
To explore alternative strategies or further model improvements, consider the K-Means model k-means or the upcoming multi-core SVM model for regression; if these two models do not meet your requirements, you can wait for the SVM or lightGBM model that I will soon develop.
This presentation aims to elucidate the strategic, technical, and analytical dimensions of trading profitability, alongside updates to our indicators and model optimization strategies, including a nuanced approach to risk/reward ratios based on the accuracy of the underlying model.
¡I will be happy to answer any other questions you may have as soon as possible!.
