Backtesting

Canonical API

result = asr.backtesting.run(prices, target_weights)
print(result.summary)

The underlying engine returns BacktestResult and supports the existing BacktestSpec and CostModel objects.

What belongs in a backtest specification

• information timing;
• target-weight timing;
• execution delay;
• transaction costs;
• missing-data policy;
• rebalancing rule;
• benchmark and performance conventions.

Performance is not validation

A high in-sample metric is not sufficient. Compare specifications, test costs and delays, use chronological holdouts, and record which decisions were made before observing the evaluation period.