Backtesting software / QuantConnect

QuantConnect: backtesting workflow, data and limits

Code-first multi-asset research in Python or C#, with downloadable orders, trades, result JSON and logs documented separately.

Cloud backtesting with LEAN · Cloud · Checked 17 October 2026 · No authenticated or installed provider test

Selected workflow

See current edition terms

Cloud compute and data entitlements separate

Rules: Code

Data: Provider datasets and configurable brokerage/reality models

Stocks: Verified workflow
Forex: Verified workflow
Futures: Verified workflow
Options: Verified workflow

Cost settings: Documented; verify defaults and instrument settings

Result access/export: Documented; format and implementation below

See named documentation and current entitlement; no native export acceptance test performed.

Limits that can change a result

LEAN source availability does not make cloud compute or every dataset free. Options coverage does not guarantee every exercise/fill/margin assumption matches a broker.

Do not infer a fill model or contract lifecycle from a chart symbol alone. Historical coverage, data adjustment, sessions, time zones, commissions, position sizing and execution settings should be written down with each run. The same entry rule can produce different results under different conventions.

A test task you can reproduce

Run one order in each intended asset class, inspect fill/fee/model configuration and compare GUI timestamps with the UTC export.

This is a suggested acceptance task, not a provider test that we performed. Save source code or rule settings, the data identifier, parameters and export. If the tool cannot accept a synthetic dataset, use an explicitly identified historical slice and record what cannot be independently controlled.

Official evidence

QuantConnect asset classes — US equity, options, Forex and futures among documented asset classes; datasets and brokerage/reality models determine actual coverage.
QuantConnect backtest results — Orders/trades CSV, result JSON, logs and generated reports. GUI times can differ from UTC exports; data access/roles apply.
QuantConnect reality models — Configurable fill, fee, slippage, brokerage and margin models. A model is not evidence of actual fills or profitable execution.

What this profile does not establish

Published features do not establish current uptime, execution quality, predictive value, suitability or profitability. Trial limits and paid data can differ from the selected software licence. We have not opened accounts, purchased licences or run the named tool’s native imports/exports. The source register makes that boundary explicit.

Compare this workflow · Trace our own controlled SMA model · Read definitions