diff --git a/src/backtesting/performance/metrics.py b/src/backtesting/performance/metrics.py
index 3b0cbcd..e8e047e 100644
--- a/src/backtesting/performance/metrics.py
+++ b/src/backtesting/performance/metrics.py
@@ -54,11 +54,16 @@ def calculate_all_metrics(self,
             # Calculate period length for annualization
             if len(equity_curve) > 1:
                 if 'datetime' in equity_curve.columns:
+                    # Real timestamps: elapsed calendar time over calendar days/yr.
                     days = (equity_curve['datetime'].iloc[-1] - equity_curve['datetime'].iloc[0]).days
+                    years = max(days / 365.25, 1 / self.trading_days)
                 else:
-                    days = len(equity_curve)  # Assume daily data
+                    # No timestamps: each row is a trading bar, so annualize by
+                    # trading days per year, not calendar days. Dividing a
+                    # trading-bar count by 365.25 over-annualizes the return.
+                    bars = len(equity_curve)
+                    years = max(bars / self.trading_days, 1 / self.trading_days)
 
-                years = max(days / 365.25, 1/self.trading_days)  # Minimum one trading day
                 metrics['annualized_return'] = (1 + metrics['total_return']) ** (1/years) - 1
             else:
                 metrics['annualized_return'] = 0.0
diff --git a/tests/backtesting/test_performance.py b/tests/backtesting/test_performance.py
index fd108f7..e0ee171 100644
--- a/tests/backtesting/test_performance.py
+++ b/tests/backtesting/test_performance.py
@@ -72,6 +72,17 @@ def test_sharpe_ratio_calculation(self):
         assert metrics['sharpe_ratio'] > 10, "Sharpe ratio should be high for consistent returns"
         assert metrics['annualized_return'] > 10, "Annualized return should be very high"
 
+    def test_annualized_return_without_datetime_uses_trading_days(self):
+        """Equity curves without a datetime column are trading bars and must be
+        annualized by trading_days, not calendar days (issue #12)."""
+        # 252 trading bars (one trading year), +20% total -> ~20% annualized.
+        equity_curve = pd.DataFrame({'total': np.linspace(100000, 120000, 252)})
+
+        metrics = self.calculator.calculate_all_metrics(equity_curve, pd.DataFrame())
+
+        # Old code divided 252 bars by 365.25, over-annualizing to ~30%.
+        assert abs(metrics['annualized_return'] - 0.20) < 0.01
+
     def test_max_drawdown_calculation(self):
         """Test maximum drawdown calculation"""
         # Create equity curve with known drawdown