diff --git a/crates/ml/src/hyperopt/adapters/dqn.rs b/crates/ml/src/hyperopt/adapters/dqn.rs
index 8a9104c33..271d60d31 100644
--- a/crates/ml/src/hyperopt/adapters/dqn.rs
+++ b/crates/ml/src/hyperopt/adapters/dqn.rs
@@ -2352,10 +2352,12 @@ impl HyperparameterOptimizable for DQNTrainer {
             min_replay_size: params.batch_size * 2, // Need at least 2x batch size
             epochs: self.epochs,
             checkpoint_frequency: (self.epochs / 5).max(1), // Save 5 checkpoints per trial, min 1
-            // C4 FIX: Re-enable early stopping with adaptive plateau window.
-            // Previous issue: plateau_window=5 with 8 epochs triggered too often.
-            // Fix: adaptive window = max(epochs / 2, 3), giving enough learning time.
-            early_stopping_enabled: true,
+            // C4: Disable early stopping for hyperopt. The Sharpe-based plateau
+            // detector fires every trial because compute_epoch_financials() is
+            // deterministic (greedy argmax on fixed validation data) and the model
+            // doesn't change enough in 8 short epochs to shift any argmax decisions,
+            // so Sharpe is bit-identical across epochs → plateau at window boundary.
+            early_stopping_enabled: false,
             q_value_floor: -5.0,
             min_loss_improvement_pct: 2.0,
             plateau_window: (self.epochs / 2).max(3),