Looking at your chart, I can see what you're referring to - particularly during the uptrend around 2016-07 to 2017-01 where the model unexpectedly identifies a bearish regime despite the clearly visible upward trend in the ASW spread. This counterintuitive behavior has several potential causes: 1. **Momentum vs. Level Confusion**: The model might be mixing signals about the rate of change (momentum) with the absolute level. During that period, the uptrend might have slowed down temporarily, causing the model to classify it as bearish even though prices were still rising overall. 2. **Training Data Imbalance**: The model's understanding of "bullish" vs "bearish" regimes comes from the historical data patterns it observes. If similar patterns in your training data were predominantly followed by drops, the model might classify them as bearish even if the immediate movement is upward. 3. **Lookback Period Effects**: The Markov model is making decisions based on ...
Commentaires
Publier un commentaire