Skip to content
All library documents

Backtesting a Threshold-Based Crypto Portfolio Rebalancer

Code Lumibot

Summary

This example shows how to run a historical backtest of a cryptocurrency portfolio using a drift rebalancer and Alpaca’s backtesting data source. The described method compares holdings with target weights and trades assets that have drifted from those targets, selling relatively overweight positions and buying relatively underweight ones. The script loads strategy parameters from configuration and uses Bitcoin as the benchmark asset.

The example sets a one-month, minute-level backtest period in early 2024, enables backtest analysis, and prints the results along with filled-order details. It is a demonstration of wiring a rebalancing strategy into a backtest, not a report of performance: no return, risk, or benchmark comparison is included in the document. Its usefulness is therefore mainly methodological, and conclusions about the strategy would require examining the configured assets, drift thresholds, costs, and actual backtest output.

Key ideas

  • The example applies a drift rebalancer to a cryptocurrency portfolio with target asset weights.
  • The strategy trades assets whose portfolio weights have moved away from their targets.
  • The script runs a historical backtest using Alpaca data and uses Bitcoin as its benchmark.
  • It prints analysis output and filled-order details, but the document gives no performance results.
  • Assessing the strategy requires the configuration and backtest assumptions, including costs and drift thresholds.

Tags

CryptoPortfolio constructionBacktestingPosition sizing
Full text
# crypto_50_50.py


```py
from datetime import datetime

import pytz

from lumibot.backtesting import AlpacaBacktesting
from lumibot.components.configs_helper import ConfigsHelper
from lumibot.credentials import ALPACA_TEST_CONFIG, IS_BACKTESTING
from lumibot.entities import Asset
from lumibot.example_strategies.drift_rebalancer import DriftRebalancer
from lumibot.tools.pandas import print_full_pandas_dataframes
from lumibot.traders.debug_log_trader import DebugLogTrader

print_full_pandas_dataframes()

"""
Strategy Description

This example demonstrates using crypto with the DriftRebalancer and AlpacaBacktesting. 
It rebalances a portfolio of assets to a target weight every time the asset drifts
by a certain threshold. The strategy will sell the assets that has drifted the most and buy the
assets that has drifted the least to bring the portfolio back to the target weights.
"""


if __name__ == "__main__":

    configs_helper = ConfigsHelper(configs_folder="example_strategies")
    parameters = configs_helper.load_config("crypto_50_50_config")

    if not IS_BACKTESTING:
        print("This strategy is not meant to be run live. Please set IS_BACKTESTING to True.")
        exit()

    if not ALPACA_TEST_CONFIG:
        print("This strategy requires an ALPACA_TEST_CONFIG config file to be set.")
        exit()

    if not ALPACA_TEST_CONFIG['PAPER']:
        print(
            "Even though this is a backtest, and only uses the alpaca keys for the data source"
            "you should use paper keys."
        )
        exit()

    tzinfo = pytz.timezone('UTC')
    backtesting_start = tzinfo.localize(datetime(2024, 1, 1))
    backtesting_end = tzinfo.localize(datetime(2024, 2, 1))
    timestep = 'minute'
    auto_adjust = True
    warm_up_trading_days = 0
    refresh_cache = False

    results, strategy = DriftRebalancer.run_backtest(
        datasource_class=AlpacaBacktesting,
        backtesting_start=backtesting_start,
        backtesting_end=backtesting_end,
        minutes_before_closing=0,
        benchmark_asset=Asset("BTC", asset_type=Asset.AssetType.CRYPTO),
        analyze_backtest=True,
        parameters=parameters,

        # For seeing logs (if using DebugLogTrader, set show_progress_bar to false)
        trader_class=DebugLogTrader,
        # show_progress_bar=True,

        # AlpacaBacktesting kwargs
        timestep=timestep,
        market=parameters['market'],
        config=ALPACA_TEST_CONFIG,
        refresh_cache=refresh_cache,
        warm_up_trading_days=warm_up_trading_days,
        auto_adjust=auto_adjust,
    )

    print(results)
    trades_df = strategy.broker._trade_event_log_df  # noqa
    filled_orders = trades_df[(trades_df["status"] == "fill")]
    print(
        "\nfilled_orders:\n",
        f"{filled_orders[['time', 'symbol', 'type', 'side', 'status', 'price', 'filled_quantity', 'trade_cost']]}"
    )

```

Shown in full with attribution under the source's licence. Licence: GPL-3.0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.