Setup views: primary-target column, floor-target prune, liveness cutoff

Three follow-ups to the gate probability floor (8f41143):

- Signals table shows the starred primary target (shared primaryTarget
  helper) instead of an independently computed max-probability best,
  so Overview, Signals and ticker details agree by construction.
- Targets pinned at the 3% probability clamp floor collapse to the
  nearest one (enhance_trade_setup + backtest candidates in parity):
  floor-pinned levels are indistinguishable to the model, so farther
  ones were duplicate 3% rows inviting lottery headlines.
- get_trade_setups only returns setups re-emitted within
  LIVE_SETUP_MAX_AGE_DAYS (3): an older latest row means the daily
  scan no longer confirms the setup, and such rows otherwise surface
  forever on Overview/Signals/ticker/alerts. History endpoints keep
  full history.

Backtest on the Jul-3 snapshot is metric-identical to the gate-floor
run on all qualified stats (1089 qualified, Sharpe 2.02, CAGR +49.6%,
DD -15.8%): the prune only removes noise the gate already rejected.

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

app/services/recommendation_service.py

@@ -45,6 +45,12 @@ _MODERATE_MAX_ATR = 4.6
 # the same tolerance the chart and alerts use, so S/R is one model app-wide.
 _SR_ZONE_TOLERANCE = 0.02
 
+# Reach-probability estimates are clamped to this band; a target at the floor
+# means "the model considers it essentially unreachable" and floor-pinned
+# targets are mutually indistinguishable.
+_PROBABILITY_CLAMP_LOW = 3.0
+_PROBABILITY_CLAMP_HIGH = 95.0
+
 
 def _clamp(value: float, low: float, high: float) -> float:
     return max(low, min(high, value))
@@ -408,7 +414,7 @@ class ProbabilityEstimator:
         elif opposed:
             probability -= signal_weight * 100.0
 
-        return round(_clamp(probability, 3.0, 95.0), 2)
+        return round(_clamp(probability, _PROBABILITY_CLAMP_LOW, _PROBABILITY_CLAMP_HIGH), 2)
 
 
 signal_conflict_detector = SignalConflictDetector()
@@ -582,6 +588,26 @@ PRIMARY_TARGET_MIN_RR = 1.5
 PRIMARY_TARGET_MIN_PROBABILITY = MIN_TARGET_PROBABILITY
 
 
+def _prune_floor_pinned_targets(targets: list[dict]) -> list[dict]:
+    """Keep only the nearest target pinned at the probability clamp floor.
+
+    Floor-pinned targets are indistinguishable to the model (true probability
+    at/below the clamp), so farther ones add no information — they just fill
+    the table with duplicate "3%" rows whose inflated R:R invites lottery
+    picks. ``targets`` is distance-sorted by the generator, so the first
+    floor-pinned entry is the nearest (most reachable) representative.
+    """
+    pruned: list[dict] = []
+    seen_floor = False
+    for target in targets:
+        if float(target.get("probability", 0.0)) <= _PROBABILITY_CLAMP_LOW:
+            if seen_floor:
+                continue
+            seen_floor = True
+        pruned.append(target)
+    return pruned
+
+
 def _select_primary_target(
     targets: list[dict],
     min_rr: float = PRIMARY_TARGET_MIN_RR,
@@ -665,6 +691,9 @@ async def enhance_trade_setup(
         # Label follows from the reach-probability: high prob = Conservative.
         target["classification"] = _classify_by_probability(target["probability"])
 
+    # Collapse duplicate floor-pinned lottery targets to the nearest one.
+    targets = _prune_floor_pinned_targets(targets)
+
     # Primary target = most-likely target with real asymmetry (see
     # _select_primary_target), not the old quality-score pick that ignored
     # probability. Sync the setup's headline target/rr_ratio so the chart, gate