Backtest recipes

End-to-end recipes for the most common backtest assembly patterns. Each is a single Python file under docs/examples/, runnable as-is, and exercises a specific composition of W15 subsystems.

Recipe	What it shows
python_realistic_backtest.py	One-call venue factory; fees + funding + liquidation + rate limits wired by default
python_multi_symbol_cross_margin.py	Atomic multi-symbol on_marks + stale-mark guard for safe cross-margin walks
python_funding_aware_perp_strategy.py	Funding settlements at 8h intervals; debit/credit account equity
python_rate_limit_aware_market_maker.py	Quote loop that respects the venue's per-endpoint rate budget

When to reach for which

• Just starting a strategy → realistic_backtest as the harness; drop your signal logic into the position-open / on-mark loop.
• Multi-symbol cross account → multi_symbol_cross_margin for the on_marks pattern; otherwise you bake in a silent footgun where forgetting one symbol leaves the cross check evaluating against stale data.
• Holding-period > 8h or any leveraged perp → funding_aware so the funding-rate drag shows up in the backtest PnL.
• Market-making cadence (>100 actions/sec) → rate_limit_aware so the venue's ban policy doesn't surprise you in live.

Discoverability

Each recipe is indexed by sync_mcp_data.py and surfaces via:

• mcp__flox__get_example(topic="backtest", language="python") — topics: strategy, connector, indicator, event-handler, risk, backtest.
• mcp__flox__docs_search("venue factory") — full-text over the docs.

There is no examples_search tool. AI agents writing a backtest from scratch should call get_example first; copying a recipe and adapting beats deriving the composition from the per-subsystem API surface.