Vectorized Backtesting with Pandas: 100x Faster Than Loops

Vectorization transforms backtesting from slow loop-heavy scripts into transparent, fast, auditable pipelines. The performance gain is important, but consistency and easier validation are the bigger wins.

Core Pattern

import pandas as pd

df["ret"] = df["close"].pct_change()
df["fast"] = df["close"].rolling(20).mean()
df["slow"] = df["close"].rolling(100).mean()
df["signal"] = (df["fast"] > df["slow"]).astype(int)
df["strategy_ret"] = df["signal"].shift(1) * df["ret"]
df["equity"] = (1 + df["strategy_ret"].fillna(0)).cumprod()

Frequent Mistakes in Vectorized Tests

• Forgetting to shift signals, causing look-ahead bias.
• Applying transaction costs only on exits or only on entries.
• Mixing timezone-naive and timezone-aware timestamps.
• Ignoring NaN handling in indicator warm-up periods.

Validation Steps

• Cross-check one sample period with an event-driven backtester.
• Run sensitivity tests across indicator windows and fee assumptions.
• Inspect distribution of returns, not only final equity value.
• Store intermediate columns for reproducibility and debugging.