diff --git a/generate_build_data.py b/generate_build_data.py
new file mode 100644
index 00000000..3f039b38
--- /dev/null
+++ b/generate_build_data.py
@@ -0,0 +1,102 @@
+import argparse
+import collections
+import csv
+import glob
+import os
+import xml.etree.ElementTree
+
+
+def generate_build_data(balance_data_path):
+    abilities = {}
+    units = {}
+
+    for xml_file_path in glob.glob(os.path.join(balance_data_path, "*.xml")):
+        tree = xml.etree.ElementTree.parse(xml_file_path)
+        root = tree.getroot()
+
+        for ability_element in root.findall("./abilities/ability"):
+            if ability_element.get("index") and ability_element.get("id"):
+                abilities[ability_element.get("index")] = ability_element.get("id")
+
+        unit_name = root.get("id")
+
+        meta_element = root.find("./meta")
+
+        if unit_name and meta_element is not None and meta_element.get("index"):
+            units[meta_element.get("index")] = unit_name
+
+        build_unit_element = root.find("./builds/unit")
+        if build_unit_element is not None:
+            build_ability_index = build_unit_element.get("ability")
+
+            if unit_name == "SCV":
+                build_ability_name = "TerranBuild"
+            elif unit_name == "Probe":
+                build_ability_name = "ProtossBuild"
+            elif unit_name == "Drone":
+                build_ability_name = "ZergBuild"
+            else:
+                build_ability_name = "{}Build".format(unit_name)
+
+            if build_ability_index:
+                abilities[build_ability_index] = build_ability_name
+
+        train_unit_elements = root.findall("./trains/unit")
+        if train_unit_elements:
+            train_ability_index = train_unit_elements[0].get("ability")
+
+            if train_ability_index:
+                abilities[train_ability_index] = "{}Train".format(unit_name)
+
+                # Handle cases where a unit can train other units using multiple ability indices.
+                # The Nexus is currently the only known example.
+                for element in train_unit_elements[1:]:
+                    element_ability_index = element.get("ability")
+                    trained_unit_name = element.get("id")
+
+                    if element_ability_index != train_ability_index and trained_unit_name:
+                        train_ability_index = element_ability_index
+
+                        abilities[train_ability_index] = "{}Train{}".format(unit_name, trained_unit_name)
+
+        research_upgrade_element = root.find("./researches/upgrade")
+        if research_upgrade_element is not None:
+            research_ability_index = research_upgrade_element.get("ability")
+            research_ability_name = "{}Research".format(unit_name)
+
+            abilities[research_ability_index] = research_ability_name
+
+    sorted_abilities = collections.OrderedDict(sorted(abilities.items(), key=lambda x: int(x[0])))
+    sorted_units = collections.OrderedDict(sorted(units.items(), key=lambda x: int(x[0])))
+
+    return sorted_units, sorted_abilities
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Generate [BUILD_VERSION]_abilities.csv and [BUILD_VERSION]_units.csv'
+                                                 ' files from exported balance data.')
+    parser.add_argument('build_version', metavar='BUILD_VERSION', type=int,
+                        help='the build version of the balance data export')
+    parser.add_argument('balance_data_path', metavar='BALANCE_DATA_PATH', type=str,
+                        help='the path to the balance data export')
+
+    args = parser.parse_args()
+
+    units, abilities = generate_build_data(args.balance_data_path)
+
+    if not units or not abilities:
+        raise ValueError("No balance data found at provided balance data path.")
+
+    with open('{}_units.csv'.format(args.build_version), 'w', newline='') as csvfile:
+        csv_writer = csv.writer(csvfile, delimiter=',')
+        for unit_index, unit_name in units.items():
+            csv_writer.writerow([unit_index, unit_name])
+
+    with open('{}_abilities.csv'.format(args.build_version), 'w', newline='') as csvfile:
+        csv_writer = csv.writer(csvfile, delimiter=',')
+        for ability_index, ability_name in abilities.items():
+            csv_writer.writerow([ability_index, ability_name])
+
+
+if __name__ == "__main__":
+    main()