AI Types

Regression Strategy

A regression strategy in bot trading, viewed from the perspective of Machine Learning (ML), involves developing a model that learns from historical market data and makes predictions about future price movements. Here, regression analysis methods are applied to model the relationships between different influencing factors (e.g., technical indicators) and the prices of an asset.

How it works in Machine Learning:

Data Collection and Preparation

The bot aggregates historical data (such as price, trading volume, technical indicators) over a defined period. This data serves as features (input variables), while the target variable is typically the future price or price change.

Training the ML Model

Using Machine Learning methods (e.g., Support Vector Machines (SVM), neural networks, or linear regression), a model is trained. The training aims to identify patterns and relationships in the historical data. In linear regression, the goal is to determine the optimal weights for the input data (features) to predict the future price or price movement as accurately as possible.

Model Validation

After training the model, validation is performed using test data that was not used for training. This makes it possible to test the model's generalization ability and avoid overfitting, i.e., merely "memorizing" the training data. Cross-validation is a common method for this. In Maly, the question "How much to buy, how long to wait between purchases, when to sell again" is additionally answered by another ML model (Reinforcement Learning / Q-Learning). Thus, both ML models work together within this AI type.

Prediction and Decision

Once the model is well validated, the bot can use it to make a prediction for the future price based on current market data (i.e., current features). This prediction can be used as a basis for a trading decision: If the bot predicts that the price will rise, it could open a buy position.

Example Using a Simple Linear Regression Model

Features: Historical price movements, moving averages, trading volume

Target: The future price or percentage change in price

The model could then recognize that a certain moving average has frequently correlated with a price increase in the next 5 hours in the past.

After the model has learned the relationship of these variables, it is able to provide a prediction for the upcoming price movement.

Advantages of ML-Based Regression Strategy

• Automation: The bot is able to make decisions based on the models automatically, without a trader having to constantly intervene.
• Adaptability: ML models are able to continuously learn from new data, which helps adapt the strategy to changing market conditions.
• Complexity: Unlike simple statistical models, ML models can recognize deeper and more complex relationships between the data.

BolRSI Strategy

The BolRSI strategy combines two proven technical indicators: Bollinger Bands and the Relative Strength Index (RSI). This strategy uses fixed, predefined standard values and deliberately does not use Q-Learning or other reinforcement learning methods for optimization.

How the BolRSI Strategy Works:

Bollinger Bands

Bollinger Bands consist of a moving average (typically 20 periods) and two bands that are each two standard deviations above and below the average. These bands expand and contract based on market volatility and help identify overbought and oversold market conditions.

Relative Strength Index (RSI)

The RSI is a momentum indicator that measures the speed and magnitude of price movements. It moves on a scale from 0 to 100. Values below 30 indicate oversold conditions (potential buy signals), while values above 70 indicate overbought conditions (potential sell signals).

Combination of Both Indicators

The BolRSI strategy generates trading signals when both indicators provide matching signals. For example, a buy signal occurs when the price touches the lower Bollinger Band and the RSI is below 30 at the same time. A sell signal is triggered when the price touches the upper band and the RSI is above 70.

Use of Standard Values

Unlike the regression strategy, BolRSI does not use dynamic optimization through machine learning. Instead, the bot works with proven standard parameters based on decades of experience in technical trading. This makes the strategy transparent, traceable, and less prone to overfitting.

Advantages of the BolRSI Strategy

• Transparency: Trading decisions are based on clearly defined rules and are easy to understand.
• Robustness: Using standard values minimizes the risk of overfitting.
• Proven Methodology: Both indicators have been in use for decades and have proven themselves in various market conditions.
• Simplicity: The strategy is easy to understand and does not require complex ML models.