diff --git a/tradingagents/dataflows/discovery/scanners/insider_buying.py b/tradingagents/dataflows/discovery/scanners/insider_buying.py
index 838421a7..eaac64ba 100644
--- a/tradingagents/dataflows/discovery/scanners/insider_buying.py
+++ b/tradingagents/dataflows/discovery/scanners/insider_buying.py
@@ -90,6 +90,9 @@ class InsiderBuyingScanner(BaseScanner):
                 else:
                     context = f"{title} {insider_name} purchased {value_str} of {ticker}"
 
+                # Scoring: cluster buys > C-suite > dollar value
+                insider_score = value + (num_insiders * 500_000) + (1_000_000 if is_c_suite else 0)
+
                 candidates.append(
                     {
                         "ticker": ticker,
@@ -101,11 +104,13 @@ class InsiderBuyingScanner(BaseScanner):
                         "insider_title": title,
                         "transaction_value": value,
                         "num_insiders_buying": num_insiders,
+                        "insider_score": insider_score,
                     }
                 )
 
-                if len(candidates) >= self.limit:
-                    break
+            # Sort by signal quality, then limit
+            candidates.sort(key=lambda c: c.get("insider_score", 0), reverse=True)
+            candidates = candidates[: self.limit]
 
             logger.info(f"Insider buying: {len(candidates)} candidates")
             return candidates
diff --git a/tradingagents/dataflows/discovery/scanners/sector_rotation.py b/tradingagents/dataflows/discovery/scanners/sector_rotation.py
index 15cf70b1..c35ed9ee 100644
--- a/tradingagents/dataflows/discovery/scanners/sector_rotation.py
+++ b/tradingagents/dataflows/discovery/scanners/sector_rotation.py
@@ -123,18 +123,20 @@ class SectorRotationScanner(BaseScanner):
                 continue
 
         # Step 3: Only call get_ticker_info() for laggard tickers (< 2% 5d move).
-        # This dramatically reduces API calls from max_tickers down to ~20-30%.
-        candidates = []
-        for ticker, ret_5d in ticker_returns.items():
-            if ret_5d > 2.0:
-                continue  # Already moved — not a laggard
+        # Sort by most-negative return first — best laggards checked before limit.
+        laggards = [(t, r) for t, r in ticker_returns.items() if r <= 2.0]
+        laggards.sort(key=lambda x: x[1])
 
-            if len(candidates) >= self.limit:
+        candidates = []
+        api_calls = 0
+        max_api_calls = self.limit * 3  # budget for get_ticker_info calls
+        for ticker, ret_5d in laggards:
+            if len(candidates) >= self.limit or api_calls >= max_api_calls:
                 break
 
+            api_calls += 1
             result = self._check_sector_laggard(ticker, accelerating_sectors, get_ticker_info)
             if result:
-                # Overwrite ret_5d with the value we already computed
                 result["stock_5d_return"] = round(ret_5d, 2)
                 candidates.append(result)
 
diff --git a/tradingagents/dataflows/discovery/scanners/technical_breakout.py b/tradingagents/dataflows/discovery/scanners/technical_breakout.py
index 3c22134b..4dfa03c1 100644
--- a/tradingagents/dataflows/discovery/scanners/technical_breakout.py
+++ b/tradingagents/dataflows/discovery/scanners/technical_breakout.py
@@ -85,12 +85,12 @@ class TechnicalBreakoutScanner(BaseScanner):
             result = self._check_breakout(ticker, data)
             if result:
                 candidates.append(result)
-            if len(candidates) >= self.limit * 2:
-                break
 
+        # Sort by strongest breakout signal, then limit
         candidates.sort(key=lambda c: c.get("volume_multiple", 0), reverse=True)
+        candidates = candidates[: self.limit]
         logger.info(f"Technical breakouts: {len(candidates)} candidates")
-        return candidates[: self.limit]
+        return candidates
 
     def _check_breakout(self, ticker: str, data: pd.DataFrame) -> Optional[Dict[str, Any]]:
         """Check if ticker has a volume-confirmed breakout."""