diff --git a/src/arcaea_offline/external/smartrte/__init__.py b/src/arcaea_offline/external/smartrte/__init__.py
new file mode 100644
index 0000000..82c7bb8
--- /dev/null
+++ b/src/arcaea_offline/external/smartrte/__init__.py
@@ -0,0 +1 @@
+from .b30_csv import SmartRteB30CsvConverter
diff --git a/src/arcaea_offline/external/smartrte/b30_csv.py b/src/arcaea_offline/external/smartrte/b30_csv.py
new file mode 100644
index 0000000..ec0405f
--- /dev/null
+++ b/src/arcaea_offline/external/smartrte/b30_csv.py
@@ -0,0 +1,64 @@
+from sqlalchemy.orm import Session
+
+from ...models import Chart, ScoreBest
+from ...utils.rating import rating_class_to_text
+
+
+class SmartRteB30CsvConverter:
+    CSV_ROWS = [
+        "songname",
+        "songId",
+        "Difficulty",
+        "score",
+        "Perfect",
+        "criticalPerfect",
+        "Far",
+        "Lost",
+        "Constant",
+        "singlePTT",
+    ]
+
+    def __init__(
+        self,
+        session: Session,
+    ):
+        self.session = session
+
+    def rows(self) -> list:
+        csv_rows = [self.CSV_ROWS.copy()]
+
+        with self.session as session:
+            results = (
+                session.query(
+                    Chart.title,
+                    ScoreBest.song_id,
+                    ScoreBest.rating_class,
+                    ScoreBest.score,
+                    ScoreBest.pure,
+                    ScoreBest.shiny_pure,
+                    ScoreBest.far,
+                    ScoreBest.lost,
+                    Chart.constant,
+                    ScoreBest.potential,
+                )
+                .join(
+                    Chart,
+                    (Chart.song_id == ScoreBest.song_id)
+                    & (Chart.rating_class == ScoreBest.rating_class),
+                )
+                .all()
+            )
+
+            for result in results:
+                # replace the comma in song title because the target project
+                # cannot handle quoted string
+                result = list(result)
+                result[0] = result[0].replace(",", "")
+                result[2] = rating_class_to_text(result[2])
+                # divide constant to float
+                result[-2] = result[-2] / 10
+                # round potential
+                result[-1] = round(result[-1], 5)
+                csv_rows.append(result)
+
+        return csv_rows