diff --git a/stats.py b/stats.py
new file mode 100644
index 00000000..c1ded8b5
--- /dev/null
+++ b/stats.py
@@ -0,0 +1,59 @@
+# -*- coding: utf-8 -*-
+import json
+
+
+def run_left_right_analysis(characters):
+  total = 0
+  right = 0
+  left = 0
+  for data in characters.itervalues():
+    if 'decomposition' not in data or 'etymology' not in data:
+      continue
+    (decomposition, etymology) = (data['decomposition'], data['etymology'])
+    if etymology['type'] != 'pictophonetic':
+      continue
+    if decomposition[0] != u'⿰' or len(decomposition) != 3:
+      continue
+    total += 1
+    phonetic = etymology.get('phonetic')
+    if phonetic == decomposition[1]:
+      left += 1
+    if phonetic == decomposition[2]:
+      right += 1
+  print '(total, left, right):', (total, left, right)
+
+
+def run_stroke_count_analysis(characters):
+  total = 0
+  counts = [0, 0, 0]
+  phonetic_stroke_total = 0
+  semantic_stroke_total = 0
+  for data in characters.itervalues():
+    etymology = data.get('etymology', {})
+    if 'phonetic' not in etymology or 'semantic' not in etymology:
+      continue
+    (phonetic, semantic) = (etymology['phonetic'], etymology['semantic'])
+    if phonetic not in characters or semantic not in characters:
+      continue
+    total += 1
+    phonetic_strokes = len(characters[phonetic]['matches'])
+    semantic_strokes = len(characters[semantic]['matches'])
+    phonetic_stroke_total += phonetic_strokes
+    semantic_stroke_total += semantic_strokes
+    counts[cmp(phonetic_strokes, semantic_strokes) + 1] += 1
+  mean = lambda x: 1.0 * x / total
+  print '(total, counts, phonetic_mean, semantic_mean):', (
+    total, map(mean, counts),
+    mean(phonetic_stroke_total), mean(semantic_stroke_total))
+
+
+if __name__ == '__main__':
+  characters = {}
+  with open('dictionary.txt') as f:
+    for line in f.xreadlines():
+      if not line:
+        continue
+      data = json.loads(line.strip())
+      characters[data['character']] = data
+  run_left_right_analysis(characters)
+  run_stroke_count_analysis(characters)