vectorized_vs_event_driven.md

Vectorized Vs Event-Driven

Both upgraded backends return BacktestResultV2, but they answer different questions.

Vectorized

Question:

What would the account curve look like if these target exposures were applied deterministically over this market matrix?

Use it for:

• grid search;
• daily/intraday alpha research;
• many symbols and many parameter combinations;
• portfolio-level diagnostics;
• simple target-position strategies.

Primary input:

• raw signals or target units.

Primary output:

• equity, returns, positions, fees, funding, margin diagnostics.

Event-Driven

Question:

What happens to these specific orders as the bar stream evolves?

Use it for:

• limit orders;
• single-order studies;
• DCA/grid ladders;
• pair/basket entry and exit;
• order status reports.

Primary input:

• explicit OrderIntent records or a BasketSpec plus signal.

Primary output:

• same BacktestResultV2 contract, plus fills and order reports.

Practical Rule

Start with native_vectorized. Move to native_event only when execution details can change the alpha result. Use nautilus to validate a smaller set of important cases against a production-grade event engine.