From 24846eb9de199e2424ff9c2f83faa3b46dc4ec0c Mon Sep 17 00:00:00 2001
From: Roger Johansson <roger@asynkron.se>
Date: Thu, 22 Jan 2026 21:31:28 +0100
Subject: [PATCH] Align README scoring docs with Score

---
 README.md                           |  8 ++--
 cmd/quickdup/strategy_score_test.go | 68 +++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 4 deletions(-)
 create mode 100644 cmd/quickdup/strategy_score_test.go

diff --git a/README.md b/README.md
index 77f7aa7..d25fc8c 100644
--- a/README.md
+++ b/README.md
@@ -65,10 +65,10 @@ Patterns with similar structure but different actual code are filtered:
 
 1. Tokenize source lines of each occurrence
 2. Compute Jaccard similarity (intersection/union of token sets)
-3. Filter patterns below threshold (default: 50%)
-4. Score patterns: `uniqueWords + (similarity × 5)`
+3. Filter patterns below threshold (default: 75%)
+4. Score patterns: `max(uniqueWords - indentImbalance, 0) * adjustedSimilarity^3 + (patternLength / 20)`
 
-This eliminates most false positives like "all error handlers look similar structurally but have different messages." High similarity (especially 100% verbatim matches) boosts the score, surfacing the most actionable duplications first.
+Where `adjustedSimilarity = max(similarity * 2 - 1, 0)`. This eliminates most false positives like "all error handlers look similar structurally but have different messages." High similarity (especially 100% verbatim matches) is heavily rewarded, surfacing the most actionable duplications first.
 
 ### Phase 4: Output
 
@@ -133,7 +133,7 @@ quickdup -path . -ext .go -debug
 | `-min`                | `2`                 | Minimum occurrences to report                                    |
 | `-min-size`           | `3`                 | Base pattern size (lines) to start growing from                  |
 | `-max-size`           | `0`                 | Maximum pattern size to grow to (0 = no limit)                   |
-| `-min-score`          | `5`                 | Minimum score (unique words + similarity bonus)                  |
+| `-min-score`          | `5`                 | Minimum score (effective words * adjusted similarity^3 + size bonus) |
 | `-min-similarity`     | `0.75`              | Minimum token similarity between occurrences (0.0-1.0)           |
 | `-top`                | `10`                | Show top N patterns by score                                     |
 | `-select`             |                     | Show detailed output for patterns (format: `skip..limit`)        |
diff --git a/cmd/quickdup/strategy_score_test.go b/cmd/quickdup/strategy_score_test.go
new file mode 100644
index 0000000..417f20c
--- /dev/null
+++ b/cmd/quickdup/strategy_score_test.go
@@ -0,0 +1,68 @@
+package main
+
+import "testing"
+
+func TestNormalizedIndentScoreBalancedFullSimilarity(t *testing.T) {
+	strategy := &NormalizedIndentStrategy{}
+	entries := []Entry{
+		NewNormalizedIndentEntry(1, "if"),
+		NewNormalizedIndentEntry(0, "return"),
+		NewNormalizedIndentEntry(-1, "}"),
+	}
+
+	score := strategy.Score(entries, 1.0)
+	if score != 3 {
+		t.Fatalf("expected score 3, got %d", score)
+	}
+}
+
+func TestNormalizedIndentScoreImbalancePenalty(t *testing.T) {
+	strategy := &NormalizedIndentStrategy{}
+	entries := []Entry{
+		NewNormalizedIndentEntry(-1, "}"),
+		NewNormalizedIndentEntry(0, "return"),
+	}
+
+	score := strategy.Score(entries, 1.0)
+	if score != 1 {
+		t.Fatalf("expected score 1, got %d", score)
+	}
+}
+
+func TestNormalizedIndentScoreLowSimilarityDropsWordBonus(t *testing.T) {
+	strategy := &NormalizedIndentStrategy{}
+	entries := []Entry{
+		NewNormalizedIndentEntry(1, "if"),
+		NewNormalizedIndentEntry(-1, "}"),
+	}
+
+	score := strategy.Score(entries, 0.4)
+	if score != 0 {
+		t.Fatalf("expected score 0, got %d", score)
+	}
+}
+
+func TestWordIndentScoreSizeBonusAndSimilarityCurve(t *testing.T) {
+	strategy := &WordIndentStrategy{}
+	words := []string{
+		"if",
+		"for",
+		"switch",
+		"return",
+		"case",
+		"else",
+		"break",
+		"continue",
+		"goto",
+		"defer",
+	}
+	entries := make([]Entry, 0, 20)
+	for i := 0; i < 20; i++ {
+		entries = append(entries, NewWordIndentEntry(0, words[i%len(words)]))
+	}
+
+	score := strategy.Score(entries, 0.85)
+	if score != 4 {
+		t.Fatalf("expected score 4, got %d", score)
+	}
+}