Make target probability and classification distance-aware

Fixes nonsensical "Conservative @ 90%" on far targets (AJG: +39% target
labelled Conservative/90%).

- Probability = chance of touching the level within the outcome horizon
  under a driftless random walk: 2*(1 - Phi(d / (ATR*sqrt(T)))), T=30d to
  match the outcome evaluator. Distance (in ATR) dominates; strength and
  alignment modulate modestly. Far targets are now correctly low-prob.
- Classification derived from that probability (>=60 Conservative,
  40-60 Moderate, <40 Aggressive) instead of distance-rank.
- Combined with the most-likely-worthwhile primary pick, the nearest
  strong resistance becomes the Conservative primary; far levels demote
  to low-probability stretch targets.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

tests/unit/test_recommendation_service.py

@@ -159,48 +159,44 @@ def test_generate_targets_respects_direction_and_order():
     assert distances == sorted(distances)
 
 
-def test_probability_ranges_by_classification():
+def test_probability_decreases_with_distance():
+    """A far target must be far less likely than a near one — no 90% at +39%."""
     config = {
         "recommendation_signal_alignment_weight": 0.15,
         "recommendation_sr_strength_weight": 0.20,
-        "recommendation_distance_penalty_factor": 0.10,
     }
-    dimension_scores = {"technical": 70.0, "momentum": 70.0}
+    dims = {"technical": 50.0, "momentum": 50.0}  # neutral, isolate the distance term
 
-    conservative = probability_estimator.estimate_probability(
-        {
-            "classification": "Conservative",
-            "sr_strength": 80,
-            "distance_atr_multiple": 1.5,
-        },
-        dimension_scores,
-        "bullish",
-        "long",
-        config,
-    )
-    moderate = probability_estimator.estimate_probability(
-        {
-            "classification": "Moderate",
-            "sr_strength": 60,
-            "distance_atr_multiple": 3.0,
-        },
-        dimension_scores,
-        "bullish",
-        "long",
-        config,
-    )
-    aggressive = probability_estimator.estimate_probability(
-        {
-            "classification": "Aggressive",
-            "sr_strength": 40,
-            "distance_atr_multiple": 6.0,
-        },
-        dimension_scores,
-        "bullish",
-        "long",
-        config,
-    )
+    def prob(atr_multiple: float, strength: float = 50.0) -> float:
+        return probability_estimator.estimate_probability(
+            {"sr_strength": strength, "distance_atr_multiple": atr_multiple},
+            dims, None, "long", config,
+        )
 
-    assert conservative > 60
-    assert 40 <= moderate <= 70
-    assert aggressive < 50
+    near = prob(1.5)
+    mid = prob(4.0)
+    far = prob(10.0)
+
+    # Monotonic decay with distance
+    assert near > mid > far
+    # Near target is genuinely likely; a 10-ATR target is a long shot
+    assert near > 60
+    assert far < 25
+
+
+def test_far_target_not_high_probability_even_with_strong_level():
+    """The AJG case: a far target stays low-probability even at max strength."""
+    config = {"recommendation_sr_strength_weight": 0.20, "recommendation_signal_alignment_weight": 0.15}
+    # ~10 ATR away, strongest possible level, fully aligned bullish
+    p = probability_estimator.estimate_probability(
+        {"sr_strength": 100, "distance_atr_multiple": 10.0},
+        {"technical": 80.0, "momentum": 80.0}, "bullish", "long", config,
+    )
+    assert p < 40   # nowhere near 90
+
+
+def test_classify_by_probability_thresholds():
+    from app.services.recommendation_service import _classify_by_probability
+    assert _classify_by_probability(75) == "Conservative"
+    assert _classify_by_probability(50) == "Moderate"
+    assert _classify_by_probability(20) == "Aggressive"