GOLD CatBoost Clusters Gen3

GOLD CatBoost Clusters Gen3 EA operates on the CatBoost algorithm using clustering methods. A set of features is fed into the CatBoost model, constructed based on price data and moving averages (MA). These include distances between the price and MAs across different periods, as well as between the MAs themselves, which help capture deviation and trend strength. The features are normalized—by dividing by the price or MA—to make them scale-invariant. Additionally, rolling window statistics (mean, standard deviation, maximum, and minimum) are used to reflect volatility and behavior over a given period. Lagged features are also included, providing past values that help the model understand temporal dynamics. Ratio-based features, such as price-to-MA and MA-to-MA, describe relative positioning. The slope of moving averages is calculated to represent the speed of trend changes, and a compression metric—defined as the sum of distances—indicates whether the market is in a flat or trending state. Finally, binary features are added to describe the trend structure, such as sequences like MA1 > MA2 > MA3 and their reverse.