sprocket-rl-parser 1.2.70

Rocket League replay parsing and analysis.

sprocket-rl-parser

An open-source project for decompiling and analyzing Rocket League replays. It exposes a Python API that returns protobuf metadata, JSON output, and a pandas DataFrame for frame-by-frame analysis.

Install

pip install sprocket-rl-parser

Quickstart (Python)

import carball

analysis_manager = carball.analyze_replay_file("path/to/replay.replay")

# Protobuf game metadata + stats
proto_game = analysis_manager.get_protobuf_data()

# JSON-friendly dict
json_game = analysis_manager.get_json_data()

# Frame-by-frame data
data_frame = analysis_manager.get_data_frame()

If you only need replay metadata and want to avoid parsing network frames entirely:

summary_manager = carball.summarize_replay_file("path/to/replay.replay")

# Protobuf metadata only
proto_game = summary_manager.get_protobuf_data()

# JSON-friendly dict without frame data
json_game = summary_manager.get_json_data()

CLI

The package installs a carball command for one-off parsing.

carball --input path/to/replay.replay --json output.json --proto output.proto --gzip frames.gzip

Output formats

Protobuf

analysis_manager.get_protobuf_data() returns a game_pb2.Game object. This is the authoritative metadata and stats output (players, teams, events, goals, etc.). The schema lives under api/ in this repo.

JSON

analysis_manager.get_json_data() returns a JSON-compatible dict matching the protobuf schema. Use analysis_manager.write_json_out_to_file() to persist the output.

DataFrame (frame-by-frame)

analysis_manager.get_data_frame() returns a pandas DataFrame with a MultiIndex column structure:

• The index is the sequential frame number.
• Columns are tuples like (object, field) where object is game, ball, or a player name.

Example: pull ball positions and a single player's positions.

ball = data_frame["ball"][["pos_x", "pos_y", "pos_z"]]
player = data_frame["SomePlayerName"][["pos_x", "pos_y", "pos_z"]]

You can also persist the DataFrame with gzip and read it back:

import gzip
from carball.analysis.utils.pandas_manager import PandasManager

with gzip.open("frames.gzip", "wb") as f:
    analysis_manager.write_pandas_out_to_file(f)

with gzip.open("frames.gzip", "rb") as f:
    df = PandasManager.read_numpy_from_memory(f)

Advanced options

analyze_replay_file supports additional flags:

• analysis_per_goal=True to analyze each goal segment independently.
• calculate_intensive_events=True for extra stats that are more expensive to compute.
• clean=False to skip data cleanup if you want rawer frame data.

analysis_manager = carball.analyze_replay_file(
    "path/to/replay.replay",
    analysis_per_goal=False,
    calculate_intensive_events=True,
    clean=True,
)

Troubleshooting

• If you see missing or invalid data, try calculate_intensive_events=False and clean=True first.
• For reproducible analysis, make sure you are parsing the raw .replay file and not a previously decompiled JSON.
• If a replay no longer has parsable network frames after a game update, use summarize_replay_file() or decompile_replay_header_only() for header-only parsing.