Primary target: probability floor stops lottery headlines

A setup's primary target could carry a ~3% probability: the picker
chose the most likely target among those with R:R >= 1.5, and after a
run-up that pool can contain only far "lottery" levels (the near,
likely levels fail the R:R floor). The lottery target's inflated R:R
then became the setup's headline and passed the activation gate's
min_rr floor - the gate's probability check only requires a value to
exist.

Fix, no new tuning knobs: the primary must clear BOTH floors
(R:R >= 1.5 AND probability >= 20%). When nothing does, fall back to
the most likely target overall, so the headline carries an honest low
R:R and the gate rejects the setup on real numbers instead of being
gamed by an unreachable target.

Deliberately NOT pure EV-maximization (p*RR): the probability model
adds strength/alignment bonuses as flat percentage points, so EV
arithmetic would scale those bonuses by (RR+1) and systematically
favor far targets on aligned setups - the same lottery bias through
the back door.

Shared by production (enhance_trade_setup) and the backtest simulator,
so backtest comparisons stay apples-to-apples. Unit tests pin the
degenerate case; full unit suite green (497 passed).

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

app/services/recommendation_service.py

@@ -575,21 +575,40 @@ def build_recommendation_snapshot(
 
 
 PRIMARY_TARGET_MIN_RR = 1.5
+# Below this the target is a lottery ticket: under the two-barrier model a
+# fair-race 1.5:1 target sits near ~34% before drift adjustments, so 20% only
+# excludes targets the model itself considers long shots.
+PRIMARY_TARGET_MIN_PROBABILITY = 20.0
 
 
-def _select_primary_target(targets: list[dict], min_rr: float = PRIMARY_TARGET_MIN_RR) -> dict | None:
+def _select_primary_target(
+    targets: list[dict],
+    min_rr: float = PRIMARY_TARGET_MIN_RR,
+    min_probability: float = PRIMARY_TARGET_MIN_PROBABILITY,
+) -> dict | None:
     """Primary = the most LIKELY target that still offers real asymmetry.
 
-    Among targets clearing a minimal R:R floor, pick the highest probability
-    (tie-break by R:R). This fixes the old pick, which ignored probability and
-    could land on the furthest, least-likely 'lottery' level. Stronger-reward
-    levels remain in the table as stretch targets. Falls back to the highest-R:R
-    target if nothing clears the floor.
+    Among targets clearing BOTH floors (R:R >= min_rr and probability >=
+    min_probability), pick the highest probability (tie-break by R:R).
+    Stronger-reward levels remain in the table as stretch targets.
+
+    Degenerate case: after a run-up, every level with acceptable R:R can be a
+    far 'lottery' target (probability at/near the model's 3% clamp floor).
+    Previously the pick was restricted to the R:R pool, so such a lottery level
+    became the headline — its inflated R:R then sailed through the activation
+    gate's min_rr floor. Now we fall back to the most likely target overall:
+    the headline carries an honest (low) R:R and the gate rejects the setup on
+    real numbers instead of being gamed by an unreachable target.
     """
     if not targets:
         return None
 
-    worthwhile = [t for t in targets if float(t.get("rr_ratio", 0.0)) >= min_rr]
+    worthwhile = [
+        t
+        for t in targets
+        if float(t.get("rr_ratio", 0.0)) >= min_rr
+        and float(t.get("probability", 0.0)) >= min_probability
+    ]
     pool = worthwhile or targets
     return max(
         pool,

tests/unit/test_recommendation_service.py

@@ -128,6 +128,32 @@ def test_primary_target_none_when_empty():
     assert _select_primary_target([]) is None
 
 
+def test_primary_target_never_headlines_a_lottery():
+    # After a run-up: the only target clearing the R:R floor is a near-impossible
+    # far level. The primary must NOT be that lottery — fall back to the most
+    # likely target overall, so the headline R:R is honest (and the activation
+    # gate rejects the setup on min_rr instead of being gamed).
+    targets = [
+        {"price": 101.0, "rr_ratio": 0.9, "probability": 55.0},   # likely, no asymmetry
+        {"price": 140.0, "rr_ratio": 5.0, "probability": 3.0},    # asymmetric lottery
+    ]
+    primary = _select_primary_target(targets)
+    assert primary is not None
+    assert primary["price"] == 101.0
+
+
+def test_primary_target_requires_probability_floor():
+    # A worthwhile-R:R target below the probability floor loses to one that
+    # clears both floors, even at lower R:R.
+    targets = [
+        {"price": 130.0, "rr_ratio": 4.0, "probability": 12.0},   # asymmetric but unlikely
+        {"price": 112.0, "rr_ratio": 1.8, "probability": 38.0},   # clears both floors ← primary
+    ]
+    primary = _select_primary_target(targets)
+    assert primary is not None
+    assert primary["price"] == 112.0
+
+
 def test_detects_sentiment_technical_conflict():
     conflicts = signal_conflict_detector.detect_conflicts(
         dimension_scores={"technical": 72.0, "momentum": 55.0, "fundamental": 50.0},