Assessing the Limits of Technical Indicators: A Negative Result Study on RSI and SMA for Crypto Trend Prediction using Supervised Learning

The volatility of cryptocurrency markets often drives traders to rely on technical indicators such as the Relative Strength Index (RSI) and Simple Moving Average (SMA) for decision-making. This study explores the effectiveness of Supervised Machine Learning techniques in predicting short-term Bitcoin (BTC-USD) market trends based on these standard indicators. Using a dataset generated via Python spanning January 2020 to December 2025, we engineered features including Distance_SMA and RSI_Status to classify daily price movements. The workflow was implemented using k-Nearest Neighbors (kNN), Decision Tree, and Logistic Regression within the Orange Data Mining framework, validated through 10-Fold Cross-Validation. Key Findings (Negative Results): Experimental results indicate that standard RSI and SMA indicators lack significant discriminative power. All models achieved Area Under the Curve (AUC) scores of approximately 0.50 and accuracy rates hovering between 50-53% (equivalent to random guessing). Failure Analysis revealed distinct modes: Logistic Regression collapsed to a majority-class baseline (underfitting), while kNN demonstrated severe overfitting when tested on training data. These findings statistically reject the hypothesis that basic RSI and SMA values alone serve as reliable predictors for Bitcoin trends in this context, supporting the weak form of the Efficient Market Hypothesis (EMH). Supplementary Material: Source Code: Included in the file list (Python/Jupyter Notebook). Dataset: Included in the file list (CSV).