diff --git a/crates/ml-dqn/src/multi_asset.rs b/crates/ml-dqn/src/multi_asset.rs
index 794abfb8d..33338e412 100644
--- a/crates/ml-dqn/src/multi_asset.rs
+++ b/crates/ml-dqn/src/multi_asset.rs
@@ -12,7 +12,6 @@
 //! MultiAssetPortfolioTracker
 //! ├── positions: HashMap<Symbol, PortfolioTracker>
 //! ├── correlation_matrix: Array2<f64>
-//! ├── opportunity_scores: HashMap<Symbol, f64>
 //! └── portfolio_history: Vec<PortfolioSnapshot>
 //! ```
 //!
@@ -67,8 +66,6 @@ pub struct MultiAssetPortfolioTracker {
     correlation_matrix: Array2<f64>,
     /// List of all symbols in order (for indexing into correlation matrix)
     symbols: Vec<Symbol>,
-    /// Opportunity scores for symbol selection (e.g., volatility, momentum)
-    opportunity_scores: HashMap<Symbol, f64>,
     /// Portfolio value history for analytics
     portfolio_history: Vec<PortfolioSnapshot>,
     /// Trade P&L history for win rate calculation
@@ -117,14 +114,10 @@ impl MultiAssetPortfolioTracker {
         // Initialize correlation matrix as identity (uncorrelated)
         let correlation_matrix = Array2::eye(num_symbols);
 
-        // Initialize opportunity scores to zero
-        let opportunity_scores = symbols.iter().map(|sym| (sym.clone(), 0.0)).collect();
-
         Self {
             positions,
             correlation_matrix,
             symbols,
-            opportunity_scores,
             portfolio_history: Vec::new(),
             trade_pnls: Vec::new(),
             max_positions: HashMap::new(),
@@ -156,13 +149,11 @@ impl MultiAssetPortfolioTracker {
             .collect();
 
         let correlation_matrix = Array2::eye(num_symbols);
-        let opportunity_scores = symbols.iter().map(|sym| (sym.clone(), 0.0)).collect();
 
         Self {
             positions,
             correlation_matrix,
             symbols,
-            opportunity_scores,
             portfolio_history: Vec::new(),
             trade_pnls: Vec::new(),
             max_positions: HashMap::new(),
@@ -300,28 +291,6 @@ impl MultiAssetPortfolioTracker {
         variance.abs().sqrt()
     }
 
-    /// Set opportunity scores for symbol selection
-    ///
-    /// # Arguments
-    ///
-    /// * `scores` - Opportunity scores per symbol (higher = more attractive)
-    pub fn set_opportunity_scores(&mut self, scores: HashMap<Symbol, f64>) {
-        self.opportunity_scores = scores;
-    }
-
-    /// Select active symbol based on opportunity scores
-    ///
-    /// # Returns
-    ///
-    /// Symbol with highest opportunity score
-    pub fn select_active_symbol(&self) -> Symbol {
-        self.opportunity_scores
-            .iter()
-            .max_by(|(_, a), (_, b)| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
-            .map(|(sym, _)| sym.clone())
-            .unwrap_or_else(|| self.symbols[0].clone())
-    }
-
     /// Build state vector for DQN input
     ///
     /// # Format