fix: dampen agreement factor by sample size in trend confidence to prevent low-evidence inflation

Agreement of 1-2 signals was inflating confidence to paper-eligible
levels (0.575) even with low credibility sources. Added log2-based
dampener that scales agreement contribution by unique source count,
saturating at n=7. Single signals now cap at 0.39 confidence,
2 signals at 0.49 — both correctly below paper threshold (0.50).

services/aggregation/worker.py

@@ -10,6 +10,7 @@ from __future__ import annotations
 
 import json
 import logging
+import math
 import time
 import uuid as _uuid
 from dataclasses import dataclass
@@ -582,7 +583,9 @@ def compute_trend_confidence(
     Confidence is based on:
     - Number of UNIQUE source documents (not raw signal count)
     - Average extraction confidence of contributing signals
-    - Signal agreement (what fraction point the same direction)
+    - Signal agreement (what fraction point the same direction),
+      dampened by sample size so that 1-2 signals agreeing doesn't
+      inflate confidence the same way 10+ signals agreeing does
     - Contradiction penalty (high contradiction lowers confidence)
 
     Returns a value in [0, 1].
@@ -611,6 +614,12 @@ def compute_trend_confidence(
     else:
         agreement = 0.5
 
+    # Dampen agreement by sample size: 1-2 signals agreeing is far less
+    # meaningful than 7+ signals agreeing.  Uses log2(n+1)/log2(8) so the
+    # dampener saturates at 1.0 around n=7 unique sources.
+    agreement_dampener = min(1.0, math.log2(unique_sources + 1) / math.log2(8))
+    agreement *= agreement_dampener
+
     # Contradiction penalty
     contradiction_penalty = contradiction_score * 0.4
 

tests/test_paper_trading_simulation.py

@@ -426,11 +426,11 @@ class TestRecommendationDrivenOrders:
         impacts = [
             ImpactRow(
                 document_id="doc-weak-1",
-                confidence=0.20,
+                confidence=0.40,
-                novelty_score=0.1,
+                novelty_score=0.3,
-                source_credibility=0.2,
+                source_credibility=0.5,
                 sentiment="positive",
-                impact_score=0.1,
+                impact_score=0.3,
                 catalyst_type="other",
                 key_facts=["Minor update"],
                 risks=[],
@@ -438,11 +438,11 @@ class TestRecommendationDrivenOrders:
             ),
             ImpactRow(
                 document_id="doc-weak-2",
-                confidence=0.15,
+                confidence=0.35,
-                novelty_score=0.1,
+                novelty_score=0.2,
-                source_credibility=0.2,
+                source_credibility=0.4,
-                sentiment="negative",
+                sentiment="positive",
-                impact_score=0.1,
+                impact_score=0.25,
                 catalyst_type="other",
                 key_facts=["Routine filing"],
                 risks=[],