Two-lens dissent restricted to cyclicals (sector or industry): secular winners are not trimmed early

src/metrics.py

@@ -1053,6 +1053,15 @@ def candidate_fit(cand: dict, holdings: list[dict], pm: dict, risk: dict, needs:
 
 
 CYCLICAL_SECTORS = {"Industrials", "Energy", "Basic Materials", "Consumer Cyclical"}
+# sector labels hide the deep cyclicals inside Technology/Healthcare (memory chips, autos…)
+_CYCLICAL_INDUSTRY = ("semiconductor", "memory", "auto", "steel", "chemical", "oil", "gas", "mining", "metal",
+                      "homebuild", "airline", "shipping", "freight", "truck", "machinery", "construction",
+                      "lumber", "paper", "copper", "aluminum", "coal", "drilling", "equipment")
+
+
+def is_cyclical(sector, industry) -> bool:
+    sec, ind = str(sector or "").title(), str(industry or "").lower()
+    return sec in CYCLICAL_SECTORS or any(k in ind for k in _CYCLICAL_INDUSTRY)
 
 
 # ---------------- practitioner layer: asymmetry, sizing, benchmark-relative risk ----------------

src/orchestrator.py

@@ -651,7 +651,9 @@ def run_analysis(rows: list[dict], emit, context: dict | None = None, budget_s:
     sector_over = {h["symbol"]: over.get("label") for h in holdings
                    if over.get("sector") and str(h["data"].get("sector") or "") == over["sector"]}
     decisions = P.decide_actions(panel, [h["symbol"] for h in holdings],
-                                 {h["symbol"]: h["weight"] for h in holdings}, fund_syms, asym, sector_over)
+                                 {h["symbol"]: h["weight"] for h in holdings}, fund_syms, asym, sector_over,
+                                 cyclical={h["symbol"] for h in holdings
+                                           if M.is_cyclical(h["data"].get("sector"), h["data"].get("industry"))})
     cautious_book = (str((context or {}).get("risk") or "").lower() == "conservative"
                      or "under 3" in str((context or {}).get("horizon") or "").lower()
                      or "defensive" in needs)

src/personas.py

@@ -495,7 +495,8 @@ def decide_actions(panel: dict[str, dict], tickers: list[str],
                    weights: dict[str, float] | None = None,
                    funds: set[str] | None = None,
                    asym: dict[str, dict] | None = None,
-                   sector_over: dict[str, str] | None = None) -> dict[str, dict]:
+                   sector_over: dict[str, str] | None = None,
+                   cyclical: set[str] | None = None) -> dict[str, dict]:
     """{ticker: {action, conviction, dispersion, votes_low, scores{persona: conv}}}.
 
     Weight-aware: however strong the conviction, "Buy More" is never issued for a
@@ -504,6 +505,7 @@ def decide_actions(panel: dict[str, dict], tickers: list[str],
     position with only lukewarm backing tilts to Trim."""
     from .metrics import FUND_CAP, effective_name_cap
     weights = weights or {}
+    cyclical = cyclical or set()
     funds = funds or set()
     book_cap = effective_name_cap(len(tickers))
     out = {}
@@ -556,12 +558,16 @@ def decide_actions(panel: dict[str, dict], tickers: list[str],
             action = "Trim"
             size_note = (f"at {w * 100:.0f}% of the book against a {cap * 100:.0f}% cap, "
                          f"size itself is the risk — trim back toward the cap")
-        elif action in ("Keep", "Buy More") and sum(1 for v in vals if v <= 35) >= 2 and len(vals) >= 4:
-            # two-lens dissent: when two analysts see damage (≤35) a momentum-led median
-            # does not earn a hold — a 13-bagger cyclical with value 35 / risk 35 gets trimmed
+        elif (action in ("Keep", "Buy More") and sum(1 for v in vals if v <= 35) >= 2 and len(vals) >= 4
+              and t in cyclical):
+            # two-lens dissent, CYCLICALS ONLY: at a cycle peak a momentum-led median does not
+            # earn a hold when two analysts see damage (a 13-bagger memory name with value 35 /
+            # risk 35). Secular winners are deliberately left alone — a winner is not sold
+            # early on valuation grumbles while its growth and trend lenses still back it.
             action = "Trim"
             low = [k for k, v in scores.items() if v <= 35]
-            size_note = f"the {' and '.join(low)} lens{'es' if len(low) > 1 else ''} see real damage — trim rather than ride the momentum"
+            size_note = (f"cyclical at a peak — the {' and '.join(low)} lens{'es' if len(low) > 1 else ''} "
+                         f"see real damage; trim rather than ride the cycle down")
         out[t] = {"action": action, "conviction": round(med), "dispersion": round(disp),
                   "votes_low": votes_low, "n_lenses": len(vals), "scores": scores,
                   "size_note": size_note}