diff --git a/stratevo/fitness_plugins/builtin/risk_adjusted_return.py b/stratevo/fitness_plugins/builtin/risk_adjusted_return.py
new file mode 100644
index 00000000..c90c8a8c
--- /dev/null
+++ b/stratevo/fitness_plugins/builtin/risk_adjusted_return.py
@@ -0,0 +1,42 @@
+"""Risk-Adjusted Return fitness function."""
+
+from __future__ import annotations
+
+from stratevo.fitness_plugins.base import BacktestResult, BaseFitness
+
+_ZERO_TRADE_PENALTY = -1.0
+
+
+class RiskAdjustedReturnFitness(BaseFitness):
+    """Fitness based on return penalised by drawdown.
+
+    Formula::
+
+        score = annual_return / (1 + abs(max_drawdown))
+
+    A strategy with a 20% return and 10% drawdown scores higher than one
+    with a 30% return and 25% drawdown:
+
+        20% / 1.10 ≈ 0.182  >  30% / 1.25 = 0.240 … wait, let's recheck:
+        0.20 / 1.10 ≈ 0.182
+        0.30 / 1.25 = 0.240
+
+    The issue spec uses ``max_drawdown_abs`` meaning the absolute magnitude.
+    ``BacktestResult.max_drawdown`` is stored as a negative float (e.g. -0.10).
+    We take ``abs()`` so the formula is always well-defined regardless of sign
+    convention.
+    """
+
+    name: str = "risk_adjusted_return"
+    description: str = "Annual return divided by (1 + |max_drawdown|)"
+    direction: str = "maximize"
+
+    def evaluate(self, result: BacktestResult) -> float:
+        if result.total_trades == 0:
+            return _ZERO_TRADE_PENALTY
+
+        dd_abs = abs(result.max_drawdown)
+        return result.annual_return / (1.0 + dd_abs)
+
+
+FITNESS_CLASSES = [RiskAdjustedReturnFitness]
diff --git a/tests/test_fitness_risk_adjusted_return.py b/tests/test_fitness_risk_adjusted_return.py
new file mode 100644
index 00000000..ac84cb40
--- /dev/null
+++ b/tests/test_fitness_risk_adjusted_return.py
@@ -0,0 +1,83 @@
+"""Tests for RiskAdjustedReturnFitness (issue #21)."""
+
+from __future__ import annotations
+
+import pytest
+
+from stratevo.fitness_plugins.base import BacktestResult
+from stratevo.fitness_plugins.builtin.risk_adjusted_return import RiskAdjustedReturnFitness
+from stratevo.fitness_plugins.registry import FitnessRegistry
+
+
+@pytest.fixture()
+def fitness() -> RiskAdjustedReturnFitness:
+    return RiskAdjustedReturnFitness()
+
+
+def _result(**kwargs) -> BacktestResult:
+    defaults = dict(total_trades=50)
+    defaults.update(kwargs)
+    return BacktestResult(**defaults)
+
+
+class TestRiskAdjustedReturnBasics:
+    def test_name_and_direction(self, fitness: RiskAdjustedReturnFitness) -> None:
+        assert fitness.name == "risk_adjusted_return"
+        assert fitness.direction == "maximize"
+
+    def test_formula_standard(self, fitness: RiskAdjustedReturnFitness) -> None:
+        r = _result(annual_return=0.20, max_drawdown=-0.10)
+        assert fitness.evaluate(r) == pytest.approx(0.20 / 1.10)
+
+    def test_formula_positive_drawdown_convention(self, fitness: RiskAdjustedReturnFitness) -> None:
+        """max_drawdown stored as positive by some callers — abs() handles it."""
+        r = _result(annual_return=0.20, max_drawdown=0.10)
+        assert fitness.evaluate(r) == pytest.approx(0.20 / 1.10)
+
+
+class TestRiskAdjustedReturnRanking:
+    def test_drawdown_penalises_high_return(self, fitness: RiskAdjustedReturnFitness) -> None:
+        """30% return dragged down by 90% drawdown should lose to 20% / 10% DD.
+
+        20% / 10% DD → 0.20 / 1.10 ≈ 0.182
+        30% / 90% DD → 0.30 / 1.90 ≈ 0.158
+        """
+        low_dd = _result(annual_return=0.20, max_drawdown=-0.10)
+        high_dd = _result(annual_return=0.30, max_drawdown=-0.90)
+        assert fitness.evaluate(low_dd) > fitness.evaluate(high_dd)
+
+    def test_same_return_lower_drawdown_wins(self, fitness: RiskAdjustedReturnFitness) -> None:
+        same_return_low_dd = _result(annual_return=0.20, max_drawdown=-0.10)
+        same_return_high_dd = _result(annual_return=0.20, max_drawdown=-0.50)
+        assert fitness.evaluate(same_return_low_dd) > fitness.evaluate(same_return_high_dd)
+
+    def test_is_better_higher_is_better(self, fitness: RiskAdjustedReturnFitness) -> None:
+        assert fitness.is_better(0.20, 0.10) is True
+        assert fitness.is_better(0.10, 0.20) is False
+
+
+class TestRiskAdjustedReturnEdgeCases:
+    def test_zero_trades_penalty(self, fitness: RiskAdjustedReturnFitness) -> None:
+        r = BacktestResult(total_trades=0)
+        assert fitness.evaluate(r) == -1.0
+
+    def test_zero_drawdown(self, fitness: RiskAdjustedReturnFitness) -> None:
+        r = _result(annual_return=0.15, max_drawdown=0.0)
+        assert fitness.evaluate(r) == pytest.approx(0.15)
+
+    def test_negative_return(self, fitness: RiskAdjustedReturnFitness) -> None:
+        r = _result(annual_return=-0.10, max_drawdown=-0.20)
+        assert fitness.evaluate(r) == pytest.approx(-0.10 / 1.20)
+
+    def test_large_drawdown(self, fitness: RiskAdjustedReturnFitness) -> None:
+        r = _result(annual_return=0.50, max_drawdown=-0.80)
+        assert fitness.evaluate(r) == pytest.approx(0.50 / 1.80)
+
+
+class TestRiskAdjustedReturnRegistry:
+    def test_auto_discover_registers_it(self) -> None:
+        reg = FitnessRegistry()
+        reg.auto_discover()
+        assert reg.has("risk_adjusted_return")
+        fn = reg.get("risk_adjusted_return")
+        assert isinstance(fn, RiskAdjustedReturnFitness)