Why Technical Analysis Failed on Bitcoin

A Quantitative Test of the Efficient Market Hypothesis

Technical analysis has been a cornerstone of trading strategies for decades. But does it actually work on cryptocurrency markets? Our research suggests otherwise.

The Efficient Market Hypothesis

The EMH states that asset prices reflect all available information. If true, technical indicators should have no predictive power—their AUC should approximate 0.5 (random chance).

Methodology

We tested 15 popular technical indicators on Bitcoin price data from 2017-2025:

• Moving Averages (SMA, EMA)
• RSI (Relative Strength Index)
• MACD
• Bollinger Bands
• And 10 others...

from sklearn.metrics import roc_auc_score
import pandas as pd

# Calculate indicator signals
signals = calculate_technical_indicators(btc_data)

# Evaluate predictive power
for indicator in indicators:
    auc = roc_auc_score(actual_returns, signals[indicator])
    print(f"{indicator}: AUC = {auc:.4f}")

Key Findings

| Indicator | AUC Score | Result | |-----------|-----------|--------| | RSI | 0.512 | Random | | MACD | 0.498 | Random | | SMA Cross | 0.507 | Random |

Conclusion: With AUC scores averaging ~0.5, technical indicators show no significant predictive power on Bitcoin, supporting the weak-form EMH.

Full paper with statistical significance tests available on Zenodo.