VectorBT vs Backtrader: Choosing Your Backtesting Framework

Backtesting is the process of simulating a trading strategy on historical data to estimate its performance. In Python, two popular backtesting frameworks serve very different needs: VectorBT (vectorized, extremely fast) and Backtrader (event-driven, highly flexible).

VectorBT: Speed First

VectorBT treats your entire historical price series as numpy arrays and runs calculations in C-level speed via numba JIT compilation. It’s 10-100x faster than Backtrader for large parameter sweeps.

Best for:

• Parameter optimization over thousands of parameter combinations
• Long histories with millions of rows
• Strategy ideas that don’t require complex order management
• Quick prototyping and signal research

import vectorbt as vbt
import pandas as pd
import numpy as np

# Download data
btc = vbt.CCXTData.download('BTC/USDT', start='2020-01-01')

# Define simple dual moving average strategy
fast_ma = btc.close.vbt.wrapper.run_window(20, win_type='simple')
slow_ma = btc.close.vbt.wrapper.run_window(50, win_type='simple')

entries = fast_ma > slow_ma
exits = fast_ma < slow_ma

# Run backtest
pf = vbt.Portfolio.from_signals(
    btc.close,
    entries,
    exits,
    fees=0.001,  # 0.1% per trade
    slippage=0.0005
)

# Analyze
total_return = pf.total_return()
sharpe = pf.sharpe_ratio()
max_dd = pf.max_drawdown()

print(f"Total return: {total_return:.2%}")
print(f"Sharpe: {sharpe:.2f}")
print(f"Max drawdown: {max_dd:.2%}")

# Parameter sweep
param_names = ['fast_window', 'slow_window']
param_product = vbt.ParamProduct([10, 20, 30, 40, 50], [50, 100, 150, 200])
results = vbt.backtest(
    lambda params: (fast_ma > slow_ma, fast_ma < slow_ma),
    param_product
)

Backtrader: Realism First

Backtrader is event-driven: it simulates each bar (candle) and processes orders through a broker simulation. It supports live trading, multiple data feeds, and complex order types.

Best for:

• Strategies with complex order management
• Portfolio-level allocation across multiple assets
• Strategies requiring realistic fill modeling
• Live trading integration

import backtrader as bt

class SMAStrategy(bt.Strategy):
    params = (('fast', 20), ('slow', 50),)

    def __init__(self):
        self.fast_ma = bt.indicators.SMA(self.data.close, period=self.p.fast)
        self.slow_ma = bt.indicators.SMA(self.data.close, period=self.p.slow)
        self.crossover = bt.indicators.CrossOver(self.fast_ma, self.slow_ma)

    def next(self):
        if self.crossover > 0:  # Golden cross
            self.buy()
        elif self.crossover < 0:  # Death cross
            self.sell()

cerebro = bt.Cerebro()
cerebro.addstrategy(SMAStrategy)

data = bt.feeds.YahooFinanceData(dataname='AAPL',
                                  fromdate='2020-01-01',
                                  todate='2024-01-01')
cerebro.adddata(data)
cerebro.broker.setcommission(commission=0.001)
cerebro.run()
print(f'Final portfolio value: {cerebro.broker.getvalue():.2f}')

Key Differences

Which to Choose

Use VectorBT when: You’re in the research phase, testing many parameter combinations, and need speed. Great for signal discovery and factor research.

Use Backtrader when: You need realistic order handling, multi-asset portfolio simulation, or you plan to eventually run the strategy live.

Use both: Run initial screening in VectorBT; validate promising strategies in Backtrader for realism.

Common Pitfalls

VectorBT: Its simplified order model can produce optimistic results. A strategy that looks great in VectorBT may perform worse in Backtrader or live due to fill timing.

Backtrader: Slow for parameter sweeps. If you’re testing 10,000 parameter combinations, Backtrader will take days.

References

• McKinney, Python for Data Analysis, Ch. 11. O’Reilly, 2023. ISBN: 978-1-492-07231-9.
• Chan, Quantitative Trading, Ch. 1. Wiley, 2008. ISBN: 978-0-470-28488-9.
• López de Prado, Advances in Financial Machine Learning, Ch. 10. Wiley, 2018. ISBN: 978-1-119-48208-6.