diff --git a/tests/test_opportunity_evaluation_service.py b/tests/test_opportunity_evaluation_service.py
new file mode 100644
index 0000000..c75113b
--- /dev/null
+++ b/tests/test_opportunity_evaluation_service.py
@@ -0,0 +1,90 @@
+"""Opportunity historical evaluation tests."""
+
+from __future__ import annotations
+
+import json
+import tempfile
+import unittest
+from pathlib import Path
+
+from coinhunter.services import opportunity_evaluation_service
+
+
+def _rows(start_ms: int, closes: list[float]) -> list[list[float]]:
+    rows = []
+    for index, close in enumerate(closes):
+        open_time = start_ms + index * 3_600_000
+        volume = 1_000 + index * 10
+        rows.append(
+            [
+                float(open_time),
+                close * 0.99,
+                close * 1.02,
+                close * 0.98,
+                close,
+                float(volume),
+                float(open_time + 3_599_999),
+                close * volume,
+            ]
+        )
+    return rows
+
+
+class OpportunityEvaluationServiceTestCase(unittest.TestCase):
+    def test_evaluate_opportunity_dataset_scores_historical_samples(self):
+        start_ms = 1_767_225_600_000
+        dataset = {
+            "metadata": {
+                "plan": {
+                    "intervals": ["1h"],
+                    "simulation_start": "2026-01-01T04:00:00Z",
+                    "simulation_end": "2026-01-01T07:00:00Z",
+                    "simulate_days": 1,
+                }
+            },
+            "klines": {
+                "GOODUSDT": {"1h": _rows(start_ms, [100, 101, 102, 103, 104, 106, 108, 109, 110])},
+                "BADUSDT": {"1h": _rows(start_ms, [100, 99, 98, 97, 96, 95, 94, 93, 92])},
+            },
+        }
+        config = {
+            "market": {"default_quote": "USDT"},
+            "opportunity": {
+                "entry_threshold": 1.5,
+                "watch_threshold": 0.6,
+                "evaluation_horizon_hours": 2.0,
+                "evaluation_take_profit_pct": 1.0,
+                "evaluation_stop_loss_pct": 2.0,
+                "evaluation_setup_target_pct": 0.5,
+                "evaluation_lookback": 4,
+                "top_n": 2,
+            },
+        }
+
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            dataset_path = Path(tmp_dir) / "opportunity-dataset.json"
+            dataset_path.write_text(json.dumps(dataset), encoding="utf-8")
+
+            payload = opportunity_evaluation_service.evaluate_opportunity_dataset(
+                config,
+                dataset_path=str(dataset_path),
+                horizon_hours=2.0,
+                take_profit=0.01,
+                stop_loss=0.02,
+                setup_target=0.005,
+                lookback=4,
+                top_n=2,
+                max_examples=3,
+            )
+
+        self.assertEqual(payload["summary"]["symbols"], ["BADUSDT", "GOODUSDT"])
+        self.assertEqual(payload["summary"]["interval"], "1h")
+        self.assertGreater(payload["summary"]["count"], 0)
+        self.assertIn("by_action", payload)
+        self.assertIn("trade_simulation", payload)
+        self.assertEqual(payload["rules"]["research_mode"], "disabled: dataset has no point-in-time research snapshots")
+        self.assertLessEqual(len(payload["examples"]), 3)
+
+
+if __name__ == "__main__":
+    unittest.main()